The Three-Verdict Problem

Abstract

In Scotland, for hundreds of years, juries have chosen between three criminal verdicts: “guilty,” “not guilty,” and “not proven.” The “not proven” verdict’s legal meaning remains mysterious. In this article, I aim to describe and solve the problem. Applying modern ideas about standards of proof to the intellectual history of “not proven” yields eight plausible meanings for the verdict. With the extent of the problem in mind, I offer a solution. In the three-verdict system, jurors should deliver a “guilty” verdict when they believe that the accused has committed the crime and a “not guilty” verdict when they believe that the accused has not committed the crime. The “not proven” verdict is for all other states of mind. Clarifying this question matters for determining whether the verdict’s existence is just. It also offers some evidence for how the criminal standard of proof works in other legal systems.

I. Introduction

Criminal trials in most jurisdictions end in one of two ways. The accused is pronounced guilty or not guilty. A “guilty” verdict licenses criminal punishment, while a “not guilty” verdict results in acquittal, which protects the accused from being tried again for the same crime.

In Scotland, it is different. Scottish juries may choose a third verdict: “not proven.” Like “not guilty,” “not proven” is a verdict of acquittal. Scottish juries are given few instructions on how the verdicts differ, often being told simply that there are two verdicts of acquittal, and they may choose either one.¹ By offering these two verdicts, Scotland appears to be unique.²

The “not proven” verdict’s existence, commentators believe, is “a matter of historical accident rather than conscious design.”³ Originating in seventeenth-century special verdict forms, which asked the jury to decide whether elements of the charged offence were proven or not proven, “not proven” endured as a third verdict after the forms went out of use.⁴ In the intervening centuries, the verdict has been criticized as inconsistent with the principle of innocent until proven guilty, and sometimes defended as a protection for the accused from wrongful convictions.⁵ Juries’ willingness to issue it has also changed over time, once making up more than half of all verdicts of acquittal and now making up substantially less.⁶

Recently, the Scottish Parliament has proposed abolishing the verdict, inviting perspectives from the public.⁷ As part of the process, the government has commissioned reports on the “effects of the unique features of the Scottish jury system on jury reasoning and jury decision making,” and in particular, “jurors’ understandings of the not proven verdict and why might they choose this over another verdict.”⁸

The conversation about whether to retain or abolish the verdict is strikingly agnostic about the verdict’s legal meaning. Because courts do not explain the verdict’s meaning to juries, the question has shifted to trying to determine how members of the public understand it.⁹ But public perceptions of the “not proven” verdict are not necessarily good evidence of the verdict’s legal meaning. Legal commentators do not take lay perceptions of burglary or manslaughter as determining the legal requirements of those crimes. Legal concepts often have specific meanings, which means if juries take a different view of these meanings, they are applying the concepts incorrectly, rather than revealing those concepts’ meaning.¹⁰

The agnosticism is all the more curious because it was not shared in previous centuries. In the nineteenth century, a comment attributed to Lord Cockburn concluded that “not proven” means that “while the jury are not satisfied with the evidence of guilt, neither are they satisfied of the prisoner’s innocence.”¹¹ Scholars offered two narrower, opposed conceptions of this idea during that century.

First, A. V. Dicey conceived of “guilty” and “not guilty” as two extremes, like a north and south pole, with “not proven” as the globe in between. The “not guilty” verdict must, Dicey concluded, “mean more than in England, and amount to a moral acquittal: it must signify not only that the jury are uncertain whether the prisoner did or did not, e.g., commit murder, but—what is a very different matter—that they are convinced that he did not commit a murder.”¹² Because a Scottish jury, unlike an English one, always has the option to return a “not proven” verdict, a Scottish “not guilty” verdict is a statement by the jury that the accused is innocent—just as a “guilty” verdict is a statement that the accused is guilty. And for all cases in which the jury is unconvinced in either direction, “not proven” is appropriate (Figure 1).

Figure 1. Dicey conception.

An opposing view, linked to Dicey’s contemporary William Forsyth, thought of the three verdicts as having a different relationship. For Forsyth, a Scottish “not guilty” verdict does not mean that the jury is convinced that the accused is innocent. Instead, a “not guilty” verdict in Scotland means something more like what it means in England or the United States. It implies only “that the legal evidence is not sufficient to produce that degree of certainty which would justify or render safe a conviction.”¹³ The jury need not be sure of an accused’s innocence to hand down a “not guilty” verdict, says Forsyth:

[Jurors] do not thereby say that he has not committed the crime, but merely that it has not been legally proved that he has. There is therefore nothing in the verdict which need alarm the most scrupulous conscience, for it may be, and indeed ought to be, given whenever a juror is not fully and beyond all reasonable doubt satisfied of the guilt of the accused.¹⁴

If this is the meaning of the verdict of “Not Guilty,” Forsyth argues, then “there are grave objections against that of Not Proven.”¹⁵ “Not guilty” is not an opposite verdict to “guilty,” announcing that the jury is convinced of the accused’s innocence. “Not guilty” verdicts include such cases, but they also include cases in which the jury is simply unconvinced of the accused’s guilt. “Not proven” must then be reserved for cases in which the jury is especially unsure of the accused’s innocence, or especially close to thinking the accused committed the crime. “One hardly sees how [the accused] can afterwards hold up his head amongst his fellow-men, when there stands recorded against him the opinion of a jury, that the evidence respecting his guilt was so strong that they did not dare to pronounce a verdict of acquittal (Figure 2).”¹⁶

Figure 2. Forsyth conception.

“Not proven” roused Forsyth’s moral concern because he believed it was inconsistent with the principle of innocent until proven guilty. “Not proven” casts an aspersion on the accused if the “not guilty” verdict includes cases in which the jury is not sure of the accused’s innocence. If we were to plot Forsyth’s verdicts along Dicey’s spectrum, “guilty” still sits at the far end; “not guilty” begins at the close end and now travels some significant way down the path; and “not proven” is in the middle distance, between “not guilty” and “guilty.”

For the remainder of this article, I will attribute the first idea—that “not guilty” amounts to a moral acquittal—to Dicey, and the second idea—that “not guilty” is intended for a wider range of cases, including all cases in which the jury is not satisfied that the accused committed the crime—to Forsyth.¹⁷ Both Dicey and Forsyth agreed that “guilty” has a fixed meaning. It is the meanings of “not guilty” and “not proven” that are in dispute.

Courts provide little help. As the Jury Manual of Scotland describes, trial judges are quite limited in what they may instruct juries on what “not proven” means. The guidance suggests simply that there are two verdicts of acquittal, and that neither permits a retrial.¹⁸ It might be that courts recognize that the meaning of the verdict is not easy to decipher, and they do not wish to risk trial judges giving juries incorrect instructions. And this is where the “not proven” verdict’s legal meaning currently stands.

But the problem of the “not proven” verdict’s meaning is deeper than previously recognized. It is harder than simply choosing between the Dicey and Forsyth accounts. This is because Dicey and Forsyth agreed on at least the meaning of “guilty.” That was the fixed star around which their disagreement about the meanings of “not guilty” and “not proven” revolved. In our time, that star is out of orbit: there is disagreement about the meaning of the “guilty” verdict even in simpler, two-verdict systems. There are, in fact, at least four conflicting theories of what a “guilty” verdict means. These are based in (1) probability assessments, (2) full belief requirements, (3) relative plausibility judgments, and (4) probabilistic knowledge acquisition. Each of the four legal meanings of “guilty” proposed by these accounts will define “not proven” differently, whether on Dicey’s conception or Forsyth’s. There are thus at least eight possible legal meanings of “not proven.”

The legal meaning of “not proven” depends on what the other two verdicts mean. It is a three-verdict problem. In this article, my first goal is to describe the problem, and to show that it extends further than previously recognized. The question matters for Scottish criminal procedure, as well as for the ongoing debate about whether to keep the verdict in Scotland.¹⁹ My other goal is to solve it. I take it that an account should be able to maintain non-arbitrary conceptual distinctions between the three verdicts, and I argue that the clearest way to divide them is through the contemporary notion of full beliefs. In Scotland, the law intends that the jury deliver a “guilty” verdict when it believes that the accused committed the crime and deliver a “not guilty” verdict when it believes that the accused did not commit the crime. For all other cases—for example, where the jury isn’t sure, or thinks that there is some chance in either direction—the “not proven” verdict is appropriate. If “P” is the proposition that the accused committed the crime, then the jury should deliver a “guilty” verdict if it thinks P is true, a “not guilty” verdict if it thinks P is false, and a “not proven” verdict if it has neither belief.²⁰ This is the best construction of the law’s intent.

My argument is thus that a combination of Dicey’s account of the Scottish “not guilty” verdict and the modern-day full belief account of the “guilty” verdict is the best explanation of the “not proven” verdict’s meaning. In two-verdict systems, by contrast—such as England or the United States—the “not guilty” verdict includes any jury belief state other than the belief that the accused committed the crime. In all these jurisdictions, the law gives the same effect—acquittal—to any jury belief state other than the belief that the accused committed the crime, but Scotland allows the jury to be more specific.

Just as theories of the criminal standard of proof are instructive for the legal meaning of “not proven,” so observing how the criminal standard of proof works with a “not proven” verdict might reveal something about how it works without one. In two-verdict systems like England and the United States, there is ongoing debate about the meaning of the guilty verdict. If the full belief account accurately describes the Scottish guilty verdict, then this is some evidence that it also accurately describes the guilty verdict in England and the United States. Of course, it is not dispositive evidence—each system could deal with the beyond a reasonable doubt standard differently. But simplicity would indicate that the full belief account correctly describes the beyond a reasonable doubt standard in general.

In their abstraction, contemporary debates over the meaning of criminal guilt have rarely made their way into judges’ opinions or instructions. Still, it is hard to avoid the sense that the subject matters outside of the university spires. Judges in different jurisdictions have tinkered over the centuries with how they explain what proof “beyond a reasonable doubt” means.²¹ Sometimes their explanations have tracked prevailing scholarly theories. It is important to the participants in a criminal justice system to know what the courts are trying to do.

If the “not proven” verdict has a specific legal meaning, then members of the public’s views on the verdict take on a different normative valence. Rather than revealing the verdict’s function, surveys of mock jurors in Scotland—as well as actual patterns of criminal verdicts—might be revealing that juries are misapplying the verdict. This question therefore matters for the ongoing conversation about whether the verdict’s existence is just. Courts’ unwillingness to describe the verdict to jurors might be allowing for widespread misapplications of a legal standard—and indeed, they might have been doing so for hundreds of years. If that is the case, then those who seek to abolish the verdict have a new theory of history for how it came to be unjustly applied, and those who seek to retain the verdict might have a more specific position to defend.

This article will (1) assess in more detail the prevailing accounts of the “not proven” verdict; (2) analyze how current theories of criminal guilt would classify these accounts; and (3) attempt to solve the problem.

II. Prevailing Understandings of “Not Proven”

The origins of the “not proven” verdict are well traversed. In the seventeenth century, Scottish indictments itemized each fact relevant to the criminal charge, and juries were expected to determine whether the alleged facts were proved.²² But because new facts could come to light during trials, and because of repeated legal substance disputes over whether the alleged facts constituted a crime, judges began issuing special interlocutors of relevancy—special verdict forms—that asked the jury to determine whether a set of facts was proven or not proven, with the ultimate finding of guilt being left to the judge.²³ After the special interlocutors system came to an end in the eighteenth century, “not proven” survived as a third verdict.

The verdict’s meaning is not straightforward, in part because Scottish appeals courts do not permit trial judges to tell jurors what the verdict means. Jury directions on this score are described in the Judicial Institute for Scotland’s Jury Manual.

Except for exceptional circumstances, the appropriate standard direction is that there are three verdicts open to them: guilty, not guilty and not proven … It is not necessary to inform the jury specifically that “not proven” is a verdict of acquittal and that the accused cannot be tried again for the same offence, but it is thought to be good practice to do so.²⁴

But it is “dangerous,” the Manual continues, “to attempt to explain any difference between the not proven and not guilty verdicts.”²⁵ Judges are advised to say just that “[n]ot guilty and not proven are verdicts of acquittal and have the same effect.”²⁶

In the context of this enforced vagueness, public discussions have focused on the verdict as it radiates outward from the bench: what jurors, trial participants, and the public at large understand it to mean.²⁷ Advocates for the verdict have argued that it communicates something to trial participants and the public, that it protects against wrongful convictions, and that it satisfies jurors. Others dispute these claims, arguing that the verdict undermines the “innocent until proven guilty” rule²⁸ or, conversely, that it makes acquittals more likely and so is too friendly to the accused.²⁹

What jurors, trial participants, and the public take the verdict to mean is not necessarily the same thing as what the verdict means as a matter of law, however.³⁰ This is true also for the historical pattern of jury verdicts. Juries used to award “not proven” verdicts more frequently than they do today,³¹ and one might take this as evidence that the verdict’s legal meaning has changed over time. But in this article I will not do so. Jurors might deliver verdicts that fail to follow the “not proven” verdict’s legal meaning, especially given how little they are told about the verdict. I propose to ask what the best construction, conceptually, of the three verdicts is. I will assume that the “not proven” verdict’s true legal meaning is to be found in the concepts embodied in the standards themselves and in their mutual relation. I take this to be the same question that interested Dicey and Forsyth.³²

Contemporary judges’ unwillingness to describe the verdict’s legal meaning have led some to conclude that the verdict simply has no distinct meaning—there are just two verdicts of acquittal, the jury can choose between them, and that is all that can be said.³³

Yet the fact that “not proven” and “not guilty” have the same official result is not enough to settle the matter of their legal meaning. Defenses to crimes like self-defense or insanity both have the same result. But self-defense—using justified force to protect oneself from another’s violence—means something different from insanity—suffering from a mental illness that makes it impossible to appreciate the wrongfulness of one’s actions and conform them to the law. Both negate criminal liability; each has a different meaning.

In the three-verdict system, “not proven” and “not guilty” have the same result. But many people think that “not proven” must lie somewhere in between “not guilty” and “guilty” in denoting the jury’s opinion of the facts. That is why, although the court of appeal tells trial judges not to explain the “not proven” verdict to jurors, trial judges have been allowed to tell jurors that “not proven” may be appropriate when there are “lingering doubts” about the accused’s guilt.³⁴

Dicey says “not guilty” verdicts have a different meaning in Scotland than in England: a moral acquittal, an assertion that the accused is innocent. Forsyth says that “not guilty” verdicts have the same meaning in both countries: the jury finds the evidence insufficient to convict the accused.

The question, I will now argue, is deepened by debates about how “guilty” verdicts work in simple two-verdict systems. For Dicey and Forsyth, “guilty” had a fixed, agreed-upon meaning, and the contested meanings of “not guilty” and “not proven” revolved around it. But for contemporary scholars, there is no consensus about what a “guilty” verdict means. There are, in fact, at least four theories about what the law intends for “guilty” verdicts to mean. Each theory, as we will see below, would categorize a “not proven” verdict in different ways. Applying Dicey’s and Forsyth’s “not guilty” verdicts to the four modern theories of “guilty” verdicts yields different results in each case.

III. Ways of Thinking About “Not Proven”

Although there have been many ideas of how the burden of proof works to sustain a “guilty” verdict, there are currently four leading theories. Each theory, taken seriously, conflicts with the other three.

The first theory is about probability. It envisions the jury’s task as assessing the likelihood that the accused committed the crime. “Beyond a reasonable doubt” means “above some probability threshold,” which, though not defined specifically by the law, might be 80 percent, 90 percent, or 99 percent. This we may call the probability account.

A second account—the full belief account—asserts that jurors may only pronounce the accused “guilty” if, after going through the evidence, they fully believe that the accused committed the crime. This is to say that they assess the probability that the accused committed the crime as a moral certainty, and have no conscious probability estimate in their minds that is less than 100 percent.

A third account—the relative plausibility account—sees the jury as comparing two stories to see which best explains the evidence: the story offered by the prosecutor or the one offered by the accused.

A fourth account—the probabilistic knowledge account—is in some senses like the probability account because it says the “guilty” verdict requires some probability threshold below 100 percent. But it asks jurors to adopt a particular state of mind with regard to the percentage. They cannot merely think that there is a high chance that the accused committed the crime; they must know that there is a high chance.

While the theories have interested law professors and philosophers, they seem to have influenced judges more rarely. There is an abstruseness to these questions that make them hard to translate into everyday legal doctrine. But I hope the following analysis will suggest that judges ought to cast an eye on these debates; depending on which of the four accounts is right, jury instructions might be describing the law incorrectly.

A. The Probability Account

On this theory, it is up to the jury to look at the facts of a case and estimate the probability that the accused has done what the prosecutor claims. This is a common way to interpret the burden of proof in two-verdict systems, especially the United States.³⁵ If the jury thinks that it is at least, say, 95 percent likely that the accused committed the crime, then it delivers a guilty verdict.

“Not proven” may be plotted along this spectrum. On Dicey’s view that “not guilty” amounts to a moral acquittal, then “not guilty” should be an opposite likelihood to “guilty.” If a “guilty” verdict means a 95 percent chance or greater that the accused committed the crime, a “not guilty” verdict would mean an equivalently small chance that the accused committed the crime. Although there might be mathematical wrinkles in how to calculate “equivalently small” along a percentage line, we might for convenience say 5 percent—the same distance from zero as the “guilty” standard’s 95 is from 100. For all cases in which the jury thinks that it is between 6 percent and 94 percent likely that the accused committed the crime, “not proven” is the proper verdict (Figure 3).

Figure 3. Dicey conception—probability account.

Under the Forsyth conception, the analysis is similar. For Forsyth, the “not guilty” verdict includes a much larger number of cases. If “guilty” means a 95 percent chance or greater, “not guilty” might mean any case in which the probability falls between zero and 70 percent that the accused committed the crime, with “not proven” being reserved for cases in which there is a 71–94 percent chance that the accused committed the crime. In either case, the percentages may be ratcheted upward or downward along the scale depending on where one sets the likelihood required by “guilty” (Figure 4).

Figure 4. Forsyth conception—probability account.

Some people might find such calculations unsettling in criminal cases. But perhaps others find them inevitable. Percentage estimates, this second group might think, are unavoidable, as life is made up of relative chances—what can we really know with 100 percent certainty?—and the only way to resolve criminal cases is to proceed on that understanding. There might be reasons not to tell jurors, or the public, what the mathematics of criminal guilt are, and to leave it up to the jurors themselves to determine to their satisfaction what probability threshold “beyond a reasonable doubt” requires. But it clearly requires some probability threshold less than 100 percent, and thus the inquiry into the standard of proof can end. That is the thought.

But despite its popularity as a shorthand description of the “guilty” verdict, scholars who write about criminal burdens of proof are unconvinced that the probability model is right. A famous type of thought experiment shows why:

Imagine that twenty-five visitors are standing in an art gallery. After the gallery guard leaves the room for a break, twenty-four of the visitors work together to steal an Impressionist painting. Some disable the security systems, others watch for the guard’s return; some remove the painting from the wall, others push the painting through a window into a waiting getaway car. The twenty-fifth visitor, who just came to see the art, tries to stop the theft but is held back by several of the conspirators. The only witnesses watch from the street through the gallery’s frosted glass windows, too far away to distinguish the individual features of the participants. By the time the guard returns, the painting is gone, and all twenty-five visitors say they are innocent.³⁶

On these facts, twenty-four of the twenty-five visitors have committed theft. If a prosecutor charged one of them at random, there would be a 96 percent chance that the accused committed the crime.³⁷ And yet most scholars feel it is impossible to grant a guilty verdict for the randomly selected accused. Something is lacking as a matter of law.³⁸ The problem, moreover, does not seem to be local. The evidence in the Impressionist painting case seems insufficient whether the theft takes place in Edinburgh, Cape Town, or Chicago.³⁹ In any jurisdiction that uses the “beyond a reasonable doubt” standard, the evidence does not seem to suffice.⁴⁰

As a result, many commentators have concluded that a “guilty” verdict requires something more than juries’ belief in a high probability that the accused committed the crime. Each of the three following proposals offers a different idea of what else is required.

B. The Full Belief Account

The full belief account asserts that a “guilty” verdict requires a different belief state from jurors than a probability below 1. In order to convict the accused, jurors must hold a full belief that the accused committed the crime.⁴¹ A full belief is not a probability assessment—or at least, it is not a probability assessment of anything less than 100 percent. Philosophers have developed theories of how full beliefs can be made to work with probability assessments, offering a plausible account of how people go about their lives—and, indeed, how we must go about our lives—sometimes making probability assessments about the world and sometimes simply having beliefs about it.⁴² Proponents of the full belief account also say it forbids juries from gambling on the outcomes of criminal trials. A criminal justice system with a 95 percent probability threshold for conviction is in some sense telling juries to aim at a 5 percent rate of mistaken convictions, which poses questions as a matter of justice.⁴³

While the ways lawyers, scholars, and the public use words and concepts like “belief” and “confidence” change over time, the idea that criminal guilt requires what we would currently conceive of as a full belief is an old one.⁴⁴ Consider this instruction in a late nineteenth-century Louisiana manslaughter case:⁴⁵

It is not sufficient you should believe his guilt only probable. In fact, no degree of probability merely will authorize a conviction; but the evidence must be of such a character and tendency as to produce a moral certainty of the prisoner’s guilt to the exclusion of reasonable doubt.⁴⁶

Here the idea of moral certainty is a separate state of mind from a belief that an accused’s guilt is merely probable.⁴⁷ And this solves the Impressionist painting case. On the full belief account, the problem with convicting a randomly selected visitor from the art gallery is that no rational juror can come to a full belief that the defendant is one of the conspirators. At best, the evidence supports a probability assessment of 96 percent, and not a full belief, that the defendant committed the crime.⁴⁸

At this point, some readers might think the full belief account imposes unrealistic requirements. One might observe that we often cannot know even the most ordinary-seeming facts about the world to a certainty of 100 percent. If I look out of the window and see that it is raining, the fact that it is raining seems quite obvious. Yet am I 100 percent sure that it is true? There is probably some chance, even if very small, that the apparent rainstorm is an elaborate hoax. Using special technology, someone could have caused it to appear to be raining when in fact it is not. Moreover, even when it is very strong, the evidence in criminal cases is rarely more convincing than the evidence of one’s own eyes. So if a guilty verdict in a criminal case requires 100 percent certainty, then few guilty verdicts could ever be awarded, just as few full beliefs could ever exist, even about things as simple as whether it is currently raining. From this line of thought, one might conclude either that the full belief account is incorrect, or maybe that it is possible to fully believe something—that it is raining, that the accused committed the crime—while assessing the chance of it as less than 1. This thought provides an opportunity to clarify the argument from full belief.

To fully believe something, people need not tell themselves that they think it is 100 percent likely. If they assign no conscious likelihood to the proposition, but simply believe that it is true, that could count as a full belief. But full belief—as I see it—really does rule out conscious probability estimates of less than 100 percent. If it does not, then the Impressionist painting case is no longer solved.⁴⁹ So how can such a requirement be realistic?

The full belief account is realistic if we take a certain, not implausible view of how people go about their lives. On this view, it is because the world is vast and uncertain that people must form full beliefs about it. One way to see this is to think about what the probability judgments we make in everyday life seem to rest upon.

To return to the rainstorm example, suppose I were to calculate the chance that the rain was really an illusion. After thinking carefully, I conclude that the possibility that a stranger is using special technology to fake the rainstorm is 0.01 percent. Thus, it is 99.99 percent likely that it is really raining. Consider all of the other things whose likelihood I would need to have calculated to come to that conclusion. There would be the question about how likely I thought it was that such technology could exist. That question would rest upon my experiences—my past. But to rest on my experiences, I would need to assess the probabilities of more basic questions: how good is my memory, to what extent is the present likely to resemble the past, what is the chance that I am undergoing a temporary madness that inhibits my judgment about the current state of rain illusion technologies, and so on? To assess the probability that it is really raining seems to require taking certain things as true, or else my calculations would take too long. Because our capacities are limited, we take things as true in order to function.⁵⁰ Our willingness to form full beliefs depends in part on the evidence available, and also on the stakes to holding such beliefs.⁵¹

If the full belief account is right, the distinction between the Dicey and Forsyth views of “not guilty” is a matter of whether the verdict requires a full belief that the accused is innocent. That is, Dicey says that “not guilty” is intended as a “guilty” verdict’s opposite. If “guilty” requires a full belief that the accused committed the crime, then “not guilty” requires a full belief that the accused did not commit the crime. “Guilty” is reserved for cases in which the jury believes with no conscious estimate less than 100 percent that the accused committed the crime, “not guilty” is for cases in which the jury believes with no conscious estimate less than 100 percent that the accused did not commit the crime, and “not proven” is for cases in which the jury has no full belief in either direction (Figure 5).

Figure 5. Dicey conception—full belief account.

An interesting implication of Dicey’s “not guilty” verdict, on the full belief account, is that purely statistical evidence, even if it shows that there is an extremely low probability that the accused committed the crime, is not enough to permit a “not guilty” verdict. We can see this if we imagine that the roles of the gallery visitors in the Impressionist painting case were reversed. Now suppose that just one of the visitors, on their own, took the painting, and the other 24 were innocent bystanders. If a prosecutor chose one of the visitors at random to be charged in this context, the odds would be quite low—4 percent—that the visitor was guilty. Nonetheless, 4 percent is above zero, so a rational juror could not form a full belief that the accused was innocent based on this evidence alone.⁵²

For Forsyth, no full belief in the accused’s innocence is needed for “not guilty.” A 100 percent chance that the accused committed the crime is necessary for a “guilty” verdict, which leaves 0–99 percent to be divided between “not guilty” and “not proven.” “Not proven” is reserved for a persistent thought in a juror’s mind of a high risk—but one that is not enough to give rise to a full belief—that the accused committed the crime. “Not guilty” will take up most potential probabilities, from 0 to some percentage, and “not proven” will take up the difference between that percentage and 100 (Figure 6).

Figure 6. Forsyth conception—full belief account.

In either case—Forsyth’s or Dicey’s—the full belief account provides a clean distinction between the “guilty” verdict and the other verdicts. Rather than percentage thresholds plotted along a scale, the “guilty” verdict is different in kind from the other two verdicts. A full belief is its own state of mind. And because the results of “guilty” and “not proven” are so dramatically different—one leads to criminal punishment, the other leads to acquittal—there are reasons to find a conceptually clean distinction between the verdicts compelling.

C. The Relative Plausibility Account

A third account of the “guilty” verdict instructs juries to reason in a comparative sense, rather than a probabilistic or belief-based one.⁵³ Juries must assess the plausibility of the prosecutor’s claim only in comparison with the account of the facts that the accused offers.

The idea is most easily applicable in the context of civil trials, in which the standard of proof is more equal between plaintiff and defendant than “beyond a reasonable doubt.” If a plaintiff is suing a defendant for breach of contract, both sides will offer an account of what happened and the jury will compare each account. But some proponents of relative plausibility also think it explains the criminal standard of proof.⁵⁴

In criminal trials, the prosecutor will offer an account of what happened: the accused stole a painting from the art museum. The accused will offer a different story: perhaps it was someone else who stole the painting, or maybe no painting was stolen at all. Both sides will have a story to tell. The two stories will limit the possible worlds for the jurors to consider to two. The jury must consider the facts as they are presented at trial, using inference-to-the-best-explanation to see which of the two stories best explains the fact presented. After having heard and viewed all the evidence, each juror will have an idea in their mind of which story is more plausible, and that will decide the case.⁵⁵

Because the “beyond a reasonable doubt” standard places a heavier burden on the prosecutor than the accused, the relative plausibility account must somehow allocate that weight, and there are two ways of doing so. On one view, the prosecutor must offer to the jury a plausible account that the accused committed the crime, and also show that no plausible account of the accused’s innocence exists.⁵⁶ To show that there is no plausible account of the accused’s innocence is not to convince the jury that there is nothing less than a 100 percent likelihood that the accused committed the crime. It is to eschew numbers altogether.

On the second view, relative plausibility theorists need not give up on numbers. Because the prosecutor’s story needs to be much more plausible than the accused’s to secure a conviction, the criminal burden of proof can be described as hypothesis testing. A “null hypothesis”—that the accused did not commit the crime—is favored unless there is significant evidence otherwise. The beyond-a-reasonable-doubt standard sets a false positive rate like 0.05—in other words, 5 percent—which accepts one wrongful conviction for every twenty innocent people who are charged, then the jury determines the likelihood ratio between the two narratives.⁵⁷

Let us consider how “not proven” might apply to each argument.

1. The Quantitative View of Relative Plausibility

The Dicey and Forsyth views of “not proven” can work with the quantitative relative plausibility account. Dicey would say that “guilty” and “not guilty” should be understood as symmetrical judgments. If “guilty” is set to accept a false positive rate of 0.05, accepting one wrongful conviction for every twenty innocent people who are charged, then “not guilty” would set the false negative rate at 0.05, accepting one wrongful acquittal for every twenty guilty people charged. This is a skewed metric, and thus “not guilty” verdicts would be presumed rare in a functioning, noncorrupt justice system. Even though “not guilty” would be symmetrical with “guilty” as a matter of theory, in a functioning justice system, prosecutors’ discernment should be such as to bring criminal charges more frequently against those who are highly likely to be guilty than those who are highly likely to be innocent, and so “not guilty” verdicts might be expected to be rarer than “guilty” ones. “Not proven” verdicts would, in any event, be the more common of the two verdicts of acquittal (Figure 7).

Figure 7. Dicey conception—relative plausibility account (quantitative).

For Forsyth, “not guilty” is not a symmetrical judgment to “guilty,” so “not guilty” would be the default verdict for most judgments of acquittal. But there would be two false positive rates for the jurors to have in their mind. The first—the “guilty” false positive rate—would be 0.05, accepting one wrongful conviction for every twenty innocent people charged. The second—the “not proven” false positive rate—would be some value above 0.05, perhaps 0.1. A 0.1 false positive rate would accept one “not proven” verdict for every ten innocent people charged. “Not proven” is considered by Forsyth to be a grey area category in between “guilty” and “not guilty.” It will include some people who committed the crime charged and others who did not (Figure 8).

Figure 8. Forsyth conception—relative plausibility account (quantitative).

On the quantitative view of relative plausibility, the Forsyth view of “not proven” may seem relatively less troubling to the presumption of innocence than it does on other accounts of the burden of proof. The more pressing moral concern—if the quantitative relative plausibility theory is correct—is that the law is comfortable with convicting innocent people, and specifically convicting one innocent person out of every twenty. It brings up the difference observed by Laurence Tribe between a system in which every jury does their best but at times makes mistakes and a system that aims at some low rate of false convictions.⁵⁸ So long as no more than one out of every twenty innocent accuseds is convicted, the second system is, by its own lights, not making mistakes. To maintain the idea that all mistaken convictions are indeed mistakes, which any justice system should try to remedy, it is hard to accept a target false positive rate above zero.

2. The Non-Quantitative View of Relative Plausibility

The relative plausibility theorists who eschew numbers can also make room for a “not proven” verdict. For these theorists, a “guilty” verdict requires the jury to conclude both that the prosecutor’s story that the accused committed the crime is plausible and that the evidence offers no plausible story that the accused did not commit the crime.⁵⁹ Thus, for Dicey, “not guilty” would require the reverse. “Not guilty” verdicts are only permissible in those cases in which the jury finds two things: (1) that the accused’s story that she did not commit the crime is plausible; and (2) that the evidence offers no plausible account that she did commit the crime (Figure 9).

Figure 9. Dicey conception—relative plausibility account (quantitative).

As with the null-hypothesis-plus-false-positive-rate account of relative plausibility, “not guilty” verdicts would be intended by the law to be extraordinarily rare. A competent prosecutor would be unlikely to bring many criminal cases in which the available evidence offers no plausible account that the accused committed the crime. Even though there is a theoretical symmetry between the two verdicts, there is a striking way in which the relative plausibility account suggests “not guilty” verdicts should be very rare in a Diceyan world.

Forsyth’s relative plausibility account is harder to describe. For Forsyth, “not guilty” is expected to take up the bulk of the verdicts of acquittal, leaving a smaller category for those cases in which the jury is unwilling to find the accused guilty, but also is unwilling to categorize the case in the larger subset of “not guilty” cases. If the non-quantifiable relative plausibility account reserves the “guilty” verdict for cases in which jurors conclude (1) that the prosecutor’s story is plausible and (2) that there is no plausible account of the evidence consistent with the accused’s innocence, then “not proven” verdicts are intended for those cases in which at least one of these factors is not satisfied. Perhaps “not proven” is for cases in which the prosecutor’s story is plausible but there is also a plausible-but-unlikely interpretation of the evidence consistent with the accused’s innocence (Figure 10).

Figure 10. Forsyth conception—relative plausibility account (quantitative).

Or perhaps “not proven” is for cases in which the prosecutor’s account is implausible, but there is also no plausible interpretation of the evidence consistent with the accused’s innocence. Maybe the jury believes that the accused committed the crime but isn’t sure how. There is more than one way to plot Forsyth’s idea of “not proven” onto the non-quantitative relative plausibility theorist’s map of guilt.

D. The Probabilistic Knowledge Account

The probabilistic knowledge account arises out of the philosophy of knowledge. This idea posits that it is possible to know probabilistic content. There is a distinction between knowing that there is a high chance of something and merely believing there to be a high chance. To attain probabilistic knowledge about some event—to know that there is a 70 percent chance that the event happened rather than to merely believe there is a 70 percent chance that it happened—requires a certain set of things to be true. Knowledge is contingent on the ability to rule out alternatives, and mere statistical evidence is not enough to produce knowledge, because it leaves open alternatives that cannot be ruled out.⁶⁰

On the probabilistic knowledge account, a guilty verdict “requires the factfinder to know some probabilistic content, namely that the likelihood that the defendant is guilty exceeds a certain threshold. This threshold is greater than .5 and in some sense less than 1.”⁶¹ If the probability threshold is below 1, then the probabilistic knowledge account is distinct from the full belief account. In this sense it is like the probability account. But in another sense, it is different. The probabilistic knowledge account maintains that there is a difference between knowing that there is a high probability that the accused committed the crime, and only believing that there is a high probability that the accused committed the crime. On the probabilistic knowledge account, a guilty verdict requires jurors to know, and not only believe, that there is a high chance that the accused committed the crime.⁶²

If the probabilistic knowledge theory of criminal guilt is correct, the Dicey and Forsyth views of the “not proven” and “not guilty” verdicts would work in the following ways. For Dicey, if criminal guilt requires that the jury know that the probability that the accused committed the crime crosses a certain threshold—for example, 95 percent—then “not guilty” would require the juror to know, and not merely believe, that the likelihood that the accused committed the crime is below a comparatively low threshold. Thus, if 95 percent or above is required for guilt, then 5 percent or below might be required for “not guilty” (Figure 11).

Figure 11. Dicey conception—probabilistic knowledge account.

Here there is an interesting symmetry with the full belief account. If Dicey has described the “not guilty” verdict correctly, and if the probabilistic knowledge theory has described the “guilty” verdict correctly, then even very low statistical probabilities would not be legally sufficient for a “not guilty” verdict. If the numbers in the Impressionist painting case were reversed, and there were 24 innocents and one guilty party, the randomly selected visitor would have just a 4 percent chance of being the person who committed the crime—a small chance, but not small enough for a “not guilty” verdict. For there is now, it seems, an alternative possibility that the jury cannot rule out: there is something about this randomly selected accused, such as their bad character, that makes them more likely to have committed the crime than an average randomly selected visitor from the gallery would be.

Forsyth’s version of “not guilty” requires no theoretical symmetry with “guilty,” so probabilistic knowledge of a very low chance that the accused committed the crime is not required. Instead, “not guilty” verdicts could be delivered in cases in which the jurors held a variety of different belief states with regard to the likelihood that the accused committed the crime. Only knowledge that there is at least a 95 percent chance that the accused committed the crime is enough for a “guilty” verdict, and anywhere from knowledge that there is a zero percent chance that the accused committed the crime to a mere belief that there is a high chance that the accused committed the crime could be categorized as “not guilty.”

For Forsyth, there must again be some category of cases that are close to but do not reach “guilty.” Maybe the law lets the jurors deliver a “not proven” verdict if they come to know that the chances are between 70 and 95 percent that the accused committed the crime (Figure 12).⁶³

Figure 12. Forsyth conception—probabilistic knowledge account.

Or maybe Forsyth would say that “not proven” is reserved for cases in which the jury does not know, but only believes, that there is some high chance that the accused committed the crime. For example, this probability could cleanly match the probabilistic knowledge requirement. If a “guilty” verdict requires the jurors to know that there is at least a 95 percent chance that the accused committed the crime, then “not proven” could require that they believe, but not know, that there is at least a 95 percent chance that the accused committed the crime. This, like the Full Belief Account, is an interesting and clean distinction between criminal guilt and receipt of a “not proven” verdict. The distinction is knowledge. But there is still an arbitrary feature to this: knowledge of a 94 percent chance generates an acquittal, while knowledge of a 95 percent chance generates a “guilty” verdict and criminal punishment.

With the extent of the problem in mind, I will now offer a solution.

IV. What “Not Proven” Means

I take it that a solution to the three-verdict problem should provide comprehensible and non-arbitrary distinctions between the verdicts. Clarity in distinguishing among concepts is its own virtue, and this virtue is perhaps especially implicated by legal concepts, given the power that such concepts can have over people’s lives. Each of the accounts that we have been considering has given a different description of the kinds of attitudes that a juror might take during a criminal trial. One account seems to give the clearest distinctions.

Imagine that there are three kinds of views a juror might take about whether the accused committed the crime.⁶⁴ First, the juror might believe that the accused committed the crime. Second, the juror might believe that the accused did not commit the crime. And third, the juror might have neither belief. The juror might, for example, feel unsure. If the first state of mind corresponds to “guilty,” the second to “not guilty,” and the third to “not proven,” then the law has a way to distinguish between the three verdicts in a conceptually non-arbitrary fashion. This explanation corresponds most clearly with Dicey’s account of “not guilty” and the modern full belief account of “guilty.”

To put it in a more familiar way to contemporary philosophers, if the proposition “P” is that the accused committed the crime, then there are three doxastic states a juror might take toward P.⁶⁵

The juror thinks P is true. A “guilty” verdict.

The juror thinks P is false. A “not guilty” verdict.

The juror thinks neither of these things. A “not proven” verdict.

The first two states are full beliefs, which means they are not categorizable as a probability assessment below 1. In both Scotland and two-verdict systems like England, the law gives the same effect to the jury belief state “P is false” and “neither.” England gives the jury just two options, whereas Scotland divides the two categories into two verdicts, each with the same legal result.

Other solutions to the problem rely on distinctions that seem either less comprehensible or more arbitrary. For example, both the probability and probabilistic knowledge accounts of criminal guilt would draw arbitrary lines between percentage thresholds: if the jury thinks (or knows) there is a 95 percent chance that the accused committed the crime,⁶⁶ then the accused is guilty and faces a full criminal punishment, whereas if the jury merely thinks (or knows) that there is a 94 percentage chance that the accused committed the crime, then the verdict is “not proven” and an acquittal. This distinction seems arbitrary. The relative plausibility account faces the same arbitrariness problem: depending on where it sets the false positive rates, the same all-or-nothing results come from very similar belief states. And Forsyth’s account of the “not guilty” verdict faces, as he points out, another arbitrariness problem: if “not guilty” already includes cases in which the jury simply thinks that it has not been legally proved that the accused has committed the crime, there are questions about what role the “not proven” verdict has to play, other than perhaps to cast a pall over the accused.

This account of the verdict’s legal meaning might not correspond with the ways juries tend to use the verdict. The verdict’s changing use patterns over time raise compelling questions.⁶⁷ If juries do not use the verdict in the way the law intends, then the legal meaning of the verdict might be a dead letter. This would be a reason for courts to change their practices of explaining little about the verdict’s legal meaning to juries.

We might briefly consider how the legal meaning that is proposed came to be. Scholars agree that the “not proven” verdict’s existence is an accident of history—a melding of two different jury systems during contingent political circumstances.⁶⁸ But the legal meaning of a standard of proof can develop over time, through use, analysis, and conversation. The background thought behind my account is that there are basic kinds of belief states that govern how people think about the world, even if they are hard to name and describe. In working through how jurists have written about the standards of proof over time, we can find evidence of which belief states are relevant to criminal verdicts. Dicey and Forsyth provide examples of this working-through of legal concepts. Modern debates about the standards of proof in two-verdict systems are another. Such concepts need not have been consciously articulated by legal decision-makers to aptly describe the law.

Finally, if criminal guilt in Scotland requires jurors to fully believe that the accused committed the crime, then this provides some evidence of how criminal guilt works in other jurisdictions. It establishes, one might say, a prima facie case for the meaning of criminal guilt in a jurisdiction such as England. This is because the jurisdictions speak with the same words—of establishing guilt beyond a reasonable doubt—and have an entangled history. They have shared a parliament for as long as the three-verdict system has existed, and a monarch for longer than that.⁶⁹ England’s history is likewise influential on the jurisdictions that inherited its common law, such as the United States.

A more serious historical investigation could certainly cast doubt on the idea that each jurisdiction’s requirements for criminal guilt are the same. But, in the absence of such contrary evidence, inquiries into the Scottish system can be treated as useful for accounts of trial proof in other systems.⁷⁰

V. Conclusion

The “not proven” verdict’s legal meaning is interesting in its own right, and it seems relevant to the conversation about whether to retain the verdict in Scotland. Does the verdict have a legal meaning, or is it up for any given jury to determine what it means? Ought courts to have explained the verdict’s meaning in more detail? Each of these questions is implicated by the idea of the verdict’s legal meaning. In the absence of court explanations, commentators must fall back on conceptual analysis.

When we shine the light of contemporary debates about the meaning of criminal guilt on the Dicey and Forsyth accounts of “not proven,” we find that there are at least eight meanings that “not proven” could have—and probably more, depending on how many shades into which these accounts can be subdivided. But, of the four theories of “guilty,” the full belief account most cleanly distinguishes between guilty and not proven, and of the two theories of “not guilty,” Dicey’s account most cleanly distinguishes between not proven and not guilty. There are three types of belief states that matter: a belief that the accused committed the crime; a belief that the accused did not commit the crime; and all other views about the question. “Guilty” applies to the first belief state, “not guilty” applies to the second, and “not proven” applies to the third.

The Scottish system also provides some evidence for the meaning of guilty verdicts in two-verdict systems like England and the United States. If the full belief account is the best way to make sense of what Scottish jurors are being asked to do in criminal trials, then—all else equal—that suggests the same might be true for its neighbor, England, and England’s common law descendant, the United States. All three jurisdictions use the same words: reasonable doubt. The simplest explanation is that all three use them in the same way.