We need hardly say that we have no wish to lessen the fairness of criminal trials. But it must be clear what fairness means in this connection. It means, or ought to mean, that the law should be such as will secure as far as possible that the result of the trial is the right one.
—Criminal Law Revision Committee1
Underlying the question of guilt or innocence is an objective truth: the defendant, in fact, did or did not commit the acts constituting the crime charged. From the time an accused is first suspected to the time the decision on guilt or innocence is made, our criminal justice system is designed to enable the trier of fact to discover the truth according to law.
—Justice Lewis Powell2
If we look closely at the criminal justice system in the United States (or almost anywhere else for that matter), it soon becomes evident that there are three distinct families of basic aims or values driving such systems. One of these core aims is to find out the truth about a crime and thus avoid false verdicts, what I will call the goal of error reduction. A second is premised on the recognition that, however much one tries to avoid them, errors will occur from time to time. This goal addresses the question of which sort of error, a false acquittal or a false conviction, is more serious, and thus more earnestly to be avoided. In short, the worry here is with how the errors distribute themselves. Since virtually everyone agrees that convicting an innocent person is a more costly mistake than acquitting a guilty one, a whole body of doctrine and practices has grown up in the common law about how to conduct trials so as to make it more likely that, when an error does occur, it will be a false acquittal rather than a false conviction. For obvious reasons, I will say that this set of issues directs itself to the question of error distribution. The third set of values driving any legal system is a more miscellaneous grab bag of concerns that do not explicitly address trial error but focus instead on other issues important to the criminal justice system. At stake here are questions about the efficient use of resources, the protection of the rights of those accused of a crime, and various other social goods, such as the sanctity of marriage (spouses cannot be made to testify against one another) or preserving good relations with other nations (diplomats cannot generally be convicted of crimes, however inculpatory the evidence). I will call these nonepistemic policy values. Such concerns will figure here because, although not grounded in the truthseeking project, their implementation frequently conflicts with the search for the truth.
Judges and legal scholars have insisted repeatedly and emphatically that the most fundamental of these values is the first: that of finding out whether an alleged crime actually occurred and, if so, who committed it. The U.S. Supreme Court put the point concisely in 1966: “The basic purpose of a trial is the determination of the truth.”3 Without ascertaining the facts about a crime, it is impossible to achieve justice, since a just resolution crucially depends on correctly figuring out who did what to whom. Truth, while no guarantee of justice, is an essential precondition for it. Public legitimacy, as much as justice, demands accuracy in verdicts. A criminal justice system that was frequently seen to convict the innocent and to acquit the guilty would fail to win the respect of, and obedience from, those it governed. It thus seems fair to say that, whatever else it is, a criminal trial is first and foremost an epistemic engine, a tool for ferreting out the truth from what will often initially be a confusing array of clues and indicators. To say that we are committed to error reduction in trials is just another way of saying that we are earnest about seeking the truth. If that is so, then it is entirely fitting to ask whether the procedures and rules that govern a trial are genuinely truth-conducive.
The effort to answer that question constitutes what, in the subtitle of this book, I have called “legal epistemology.” Applied epistemology in general is the study of whether systems of investigation that purport to be seeking the truth are well engineered to lead to true beliefs about the world. Theorists of knowledge, as epistemologists are sometimes known, routinely examine truthseeking practices like science and mathematics to find out whether they are capable of delivering the goods they seek.
Legal epistemology, by contrast, scarcely exists as a recognized area of inquiry. Despite the nearly universal acceptance of the premise that a criminal trial is a search for the truth about a crime, considerable uncertainty and confusion reign about whether the multiple rules of proof, evidence, and legal procedure that encumber a trial enhance or thwart the discovery of the truth. Worse, there has been precious little systematic study into the question of whether existing rules could be changed to enhance the likelihood that true verdicts would ensue. legal epistemology, properly conceived, involves both a) the descriptive project of determining which existing rules promote and which thwart truth seeking and b) the normative one of proposing changes in existing rules to eliminate or modify those rules that turn out to be serious obstacles to finding the truth.
The realization of a legal epistemology is made vastly more difficult because, as just noted, nonepistemic values are prominently in play as well as epistemic ones. In many but not all cases, these nonepistemic values clash with epistemic ones. Consider a vivid example. If we were serious about error reduction, and if we likewise recognized that juries sometimes reach wrong verdicts, then the obvious remedy would be to put in place a system of judicial review permitting appeals of both acquittals and convictions. We have the latter, of course, but not the former. Every erroneous acquittal eludes detection because it escapes review. The absence of a mechanism for appealing acquittals is patently not driven by a concern to find the truth; on the contrary, such an asymmetry guarantees far more errors than are necessary. The justification for disallowing appeal of acquittals hinges on a policy value. Double jeopardy, as it is known, guarantees that no citizen can be tried twice for the same crime. Permitting the appeal of an acquittal, with the possibility that the appeal would be reversed and a new trial ordered, runs afoul of the right not to be tried more than once. So, we reach a crossroads, seemingly faced with having to choose between reducing errors and respecting traditional rights of defendants. How might we think through the resolution of conflicts between values as basic as these two are? Need we assume that rights always trump the search for the truth, or vice versa? Or, is there some mechanism for accommodating both sorts of concerns? Such questions, too, must form a core part of the agenda of legal epistemology.
This book is a first stab at laying out such an agenda. In this chapter, I formulate as clearly as I can what it means to speak of legal errors. Absent a grasp of what those errors are, we obviously cannot begin to think about strategies for their reduction. In Chapters 2 through 4, we examine in detail a host of important questions about error distribution. Chapters 5 through 8 focus on existing rules of evidence and procedure that appear to pose serious obstacles to truth seeking. Those chapters include both critiques of existing rules and numerous suggestions for fixing such flaws as I can identify. The final chapter assays some possible solutions to the vexatious problems generated by the tensions between epistemic values and nonepistemic ones.
The two passages in the epigraph to this chapter from Supreme Court Justice Lewis Powell and England’s Criminal Law Revision Committee articulate a fine and noble aspiration: finding out the truth about the guilt or innocence of those suspected of committing crimes. Yet, if read as a description of the current state of American justice, they remain more an aspiration than a reality. In saying this, I do not mean simply that injustices, false verdicts, occur from time to time. Occasional mistakes are inevitable, and thus tolerable, in any form of human inquiry. I mean, rather, that many of the rules and procedures regulating criminal trials in the United States – rules for the most part purportedly designed to aid the truth-finding process – are themselves the cause of many incorrect verdicts. I mean, too, that the standard of proof relevant to criminal cases, beyond reasonable doubt, is abysmally unclear to all those – jurors, judges, and attorneys – whose task is to see that those standards are honored. In the chapters that follow, I will show that the criminal justice system now in place in the United States is not a system that anyone concerned principally with finding the truth would have deliberately designed.4
A natural way to test that hypothesis would be to examine these rules, one by one, to single out those that thwart truth seeking. And, in the chapters to follow, I will be doing a fair share of precisely that. But, as we will discover, it is often harder than it might seem to figure out whether a given evidential practice or procedure is truth promoting or truth thwarting. In short, we need some guide lines or rules of thumb for deciding whether any given legal procedure furthers or hinders epistemic ends. Moreover, for purposes of analysis, we need to be able to leave temporarily to one side questions about the role of nonepistemic values in the administration of justice. We will have to act as if truth finding were the predominant concern in any criminal proceeding. In real life, of course, that is doubtful.
As I noted at the outset, criminal trials are driven by a host of extra-epistemic values, ranging from concerns about the rights of the defendant to questions of efficiency and timeliness. (Not for nothing do we insist that justice delayed is justice denied.) The prevailing tendency among legal writers is to consider all these values – epistemic and nonepistemic – as bundled together. This, I think, can produce nothing but confusion. Instead of the familiar form of analysis, which juggles all these values in midair at the same time, I am going to propose a thought experiment. I will suggest that we focus initially entirely on questions of truth seeking and error avoidance. I will try to figure out what sorts of rules of evidence and procedure we might put in place to meet those ends and will identify when existing rules fail to promote epistemic ends. Then, with that analysis in hand, we can turn to compare the current system of evidence rules and procedures with a system that is, as it were, epistemically optimal. When we note, as we will repeatedly, discrepancies between the kind of rules we would have if truth seeking were really the basic value and those rules we find actually in place, we will be able then to ask ourselves whether these epistemically shaky rules conduce to values other than truthseeking and, if they do, when and whether those other values should prevail over more epistemically robust ones. Although I ignore such values in the first stage of the analysis, I do not mean for a moment to suggest that they are unimportant or that they can be ignored in the final analysis. But if we are to get a handle on the core epistemic issues that are at stake in a criminal trial, it is best – at the outset – to set them to one side temporarily.
If it seems madcap to try to understand the legal system by ignoring what everyone concedes to be some of its key values, I remind you that this method of conceptual abstraction and oversimplification has proved its value in other areas of intellectual activity, despite the fact that every oversimplification is a falsification of the complexities of the real world. Consider what is perhaps the best-known example of the power of this way of proceeding: During the early days of what came to be known as the scientific revolution, Galileo set out to solve a conundrum that had troubled natural philosophers for almost two millennia, to wit, how heavy bodies fall. Everyone vaguely understood that the velocity of fall was the result of several factors. The shape of a body makes a difference: A flat piece of paper falls more slowly than one wadded into a ball. The medium through which a body is falling likewise makes a crucial difference: Heavy bodies fall much faster through air than they do through water or oil. Earlier theories of free fall had identified this resistance of the medium as the key causal factor in determining the velocity of fall. Galileo’s strategy was to turn that natural assumption on its head. Let us, he reasoned, ignore the shapes of bodies and their weights and the properties of the media through which they fall – obvious facts all. Assume, he suggested, that the only relevant thing to know is how powerfully bodies are drawn to the earth by virtue of what we would now call the gravitational field in which they find themselves. By making this stark simplification of the situation, Galileo was able to develop the first coherent account of fall, still known to high school students as Galileo’s Law. Having formulated a model of how bodies would fall if the resistance of the medium were negligible (which it is not) and the shape of the body were irrelevant (which it likewise is not), and the weight of a body were irrelevant (which it is), Galileo proceeded to reinsert these factors back into the story in order to explain real-world phenomena – something that would have been impossible had he not initially ignored these real-world constraints. The power of a model of this sort is not that it gets things right the first time around, but that, having established how things would go under limited and well-defined conditions, we can then introduce further complexities as necessary, without abandoning the core insights offered by the initial abstraction.
I have a similar thought experiment in mind for the law. Taking the Supreme Court at its word when it says that the principal function of a criminal trial is to find out the truth, I want to figure out how we might conduct criminal trials supposing that their predominant aim were to find out the truth about a crime. Where we find discrepancies between real-world criminal procedures and epistemically ideal ones (and they will be legion), we will need to ask ourselves whether the epistemic costs exacted by current real-world procedures are sufficiently outweighed by benefits of efficiency or the protection of defendant rights to justify the continuation of current practices.
Those will not be easy issues to resolve, involving as they do a weighing of values often considered incommensurable. But such questions cannot even be properly posed, let alone resolved, until we have become much clearer than we now are about which features of the current legal regime pose obstacles to truth seeking and which do not. Because current American jurisprudence tends to the view that rights almost invariably trump questions of finding out the truth (when those two concerns are in conflict), there has been far less discussion than is healthy about whether certain common legal practices – whether mandated by common law traditions or by the U.S. Constitution or devised as court-designed remedies for police abuses – are intrinsically truth thwarting.
My object in designing this thought experiment is to open up conceptual space for candidly discussing such questions without immediately butting up against the purported argument stopper: “but X is a right” or “X is required (or prohibited) by the Constitution.” Just as Galileo insisted that he wouldn’t talk about the resistance of the air until he had understood how bodies would fall absent resistance, I will try – until we have on the table a model of what a disinterested pursuit of the truth in criminal affairs would look like – to adhere to the view that the less said about rights, legal traditions, and constitutional law, the better.
I said that this thought experiment will involve figuring out how criminal trials could be conducted, supposing that true verdicts were the principal aim of such proceedings. This might suggest to the wary reader that I intend to lay out a full set of rules and procedures for conducting trials, starting from epistemic scratch, as it were. That is not quite the project I have in mind here, since it is clear that there is a multiplicity of different and divergent ways of searching for the truth, which (I hasten to add) is not the same thing as saying that there are multiple, divergent truths to be found. Consider one among many questions that might face us: If our aim is to maximize the likelihood of finding the truth, should we have trial by judge or trial by jury? I do not believe that there is a correct answer to that question since it is perfectly conceivable that we could design sets of procedures that would enable either a judge or a jury to reach verdicts that were true most of the time. English speakers have a fondness for trial by jury, whereas Roman law countries prefer trial by judge or by a mixed panel of judges and jurors. For my part, I can see no overwhelming epistemic rationale for a preference for one model over the other. If we Anglo-Saxons have any rational basis, besides familiarity, for preferring trial by jury, it has more to do with the political and social virtues of a trial by one’s peers rather than with any hard evidence that juries’ verdicts are more likely to be correct than judges’ verdicts are.
To begin with, I intend to propose a series of guidelines that will tell us what we should look for in deciding whether any particular arrangement of rules of evidence and procedure is epistemically desirable. This way of proceeding does not directly generate a structure of rules and procedures for conducting trials. What it will do is tell us how to evaluate bits and pieces of any proposed structure with respect to their epistemic bona fides. It will set hurdles or standards for judging any acceptable rule of evidence or procedure. If you want an analogy, think of how the rules of proof in mathematics work. Those rules do not generally generate proofs by some sort of formal algorithm; bright mathematicians must do that for themselves. What the rules of proof do (except in very special circumstances) is enable mathematicians to figure out whether a purported proof is a real proof. In effect, what I will be suggesting is a set of meta-rules or meta-principles that will function as yardsticks for figuring out whether any given procedure or evidence-admitting or evidence-excluding practice does, in fact, further epistemic ends or whether it thwarts them.
What I am proposing, then, is, in part, a meta-epistemology of the criminal law, that is, a body of principles that will enable us to decide whether any given legal procedure or rule is likely to be truth-conducive and error reducing. The thought experiment I have been describing will involve submitting both real and hypothetical procedures to the scrutiny that these meta-principles can provide. When we discover rules currently in place that fail to serve epistemic ends, we will want to ask ourselves whether they cannot be replaced by rules more conducive to finding the truth and minimizing error. If we can find a more truth-conducive counterpart for truth-thwarting rules, we will then need to decide whether the values that the original rules serve (for instance, protecting certain rights of the accused) are sufficiently fundamental that they should be allowed to prevail over truth seeking.
If, as Justice Powell says in the epigraph, the system “is designed” to discover the truth, you might reasonably have expected that we already know a great deal about the relation of each of its component parts to that grand ambition. The harsh reality is that we know much less than we sometimes think we do. Many legal experts and appellate judges, as we will see on numerous occasions in later chapters, continue to act and write as if certain portions of the justice system that actually thwart truth seeking have an epistemic rationale. Still worse, some jurists and legal scholars attribute error-reducing power to rules and doctrines that, viewed dispassionately, produce abundant false verdicts in their own right. Like Powell, they pay lip service to the mantra that the central goal of the system is to get at the truth, all the while endorsing old rules, or putting in place new ones, that hobble the capacity of that system to generate correct verdicts. So long as jurists believe, as many now do, that certain judicial rules (for instance, the suppression of “coerced” confessions5) promote truth finding – when in fact they do the opposite – there can be nothing but confusion concerning when and if truth seeking is being furthered.
One important reason that we know so much less than we should is that the courts in particular, but also the justice system in general, tend to discourage the sort of empirical research that would enable us to settle such questions definitively. In philosophy, my biases lean in the direction of naturalism. That means that I believe that most philosophical issues ultimately hinge on finding out what the facts are. I believe, further, that our methods of inquiry must be constantly reviewed empirically to see whether they are achieving what we expect of them. In writing this book, I have been constantly frustrated by the paucity of empirical information that would allow us to reach clear conclusions about how well or badly our legal methods are working. Where there are reliable empirical studies with a bearing on the issues addressed here, I will make use of them. Unfortunately, given the dearth of hard evidence, the analysis in this book will fall back on armchair hunches about the likely effects of various rules and procedures far more often than I would have liked. My defense for doing so is simply that one must fight one’s battles with the weapons that one has at hand.
I should stress, as well, that I approach these questions as a philosopher, looking at the law from the outside, rather than as an attorney, working within the system. Although I have thought seriously about these issues over several years, I cannot possibly bring to them the competences and sensibilities of a working trial lawyer.6 What interests me about the law is the way in which it functions, or malfunctions, theoretically, as a system for finding truth and avoiding error. In this role, I am less concerned than a civil libertarian or defense attorney might be with the rights of the accused and more concerned with how effectively the criminal justice system produces true verdicts. The analysis offered in this book does not purport to tell juries and judges how to decide a case; such dreadful decisions must depend on the case’s special circumstances and its nuances. Its aim, rather, is the more prophylactic one of pointing out some errors that these fact finders should avoid in the always difficult quest for a true and just verdict.
There will be readers who expect any avowedly philosophical treatment of the law to center on issues of morality and rights or on questions about the authority and essence of the law. Such are the themes that have dominated the philosophy of law in the last half-century. The most influential philosopher of law in the English-speaking world in the twentieth century, H. L. A. Hart, managed to write a lengthy, splendid book on the philosophy of law (The Concept of Law, 1961) that says virtually nothing about what I am calling legal epistemology. His eminent continental counterpart, Hans Kelsen, did virtually the same thing a generation earlier in his Pure Theory of Law (1934). Readers expecting a similar agenda from me will be sorely disappointed. To them in particular, I say this: If it is legitimate and fruitful for moral philosophers, such as Gerald Dworkin or John Rawls, to focus on the law principally as an exercise in ethics and morality, while largely ignoring the importance of truth seeking in the law (which they famously do), it is surely just as appropriate to look at the law through the lenses of epistemology and the theory of knowledge. Although one is not apt to learn so by looking at the existing philosophical literature on the subject, it is indisputable that the aims of the law, particularly the criminal law, are tied to epistemic concerns at least as profoundly as they are to moral and political ones. This book is a deliberate shot across the bow of the juggernaut that supposes that all or most of the interesting philosophical puzzles about the law concern its moral foundations or the sources of its authority.
In this initial chapter, I will to begin to lay out some of the analytic tools that we will need in order to grapple with some thorny problems in the theory and practice of the criminal law. As its title already makes clear, this book is largely about legal errors. Since treating the law as an exercise in epistemology inevitably means that we will be involved in diagnosing the causes of error, we need to be clear from the outset about what kinds of errors can occur in a criminal proceeding.
Since our concern will be with purely epistemic errors, I should say straight away that I am not using the term “error” as appellate courts are apt to use it. For them, an “error” occurs in a trial just in case some rule of evidence or procedure has been violated, misinterpreted, or misapplied. Thus, a higher court may determine that an error occurred when a trial judge permitted the introduction of evidence that the prevailing rules should have excluded or when some constitutional right of the defendant was violated. Courts will find that an error occurred if a judge, in his instructions to the jury about the law, made some serious mistake or other, in the sense of characterizing the relevant law in a way that higher courts find misleading or incorrect. Very occasionally, they will decide that an error occurred if the jury convicted someone when the case against the defendant failed to meet the standard of proof beyond a reasonable doubt.7
By contrast, I will be using the term “error” in a more strictly logical and epistemic sense. When I say that an error has occurred, I will mean either a) that, in a case that has reached the trial stage and gone to a verdict, the verdict is false, or b) that, in a case that does not progress that far, a guilty party has escaped trial or an innocent person has pleaded guilty and the courts have accepted that plea. In short, for the purposes of our discussion, an error occurs when an innocent person is deemed guilty or when a guilty person fails to be found guilty. For obvious reasons, I will call the first sort of error a false inculpatory finding and the second a false exculpatory finding.
There are two important points to note about the way in which I am defining legal errors:
First, errors, in my sense, have nothing to do with whether the system followed the rules (the sense of “error” relevant for appellate courts) and everything to do with whether judicial outcomes convict the guilty and free the innocent. Even if no errors of the procedural sort that worries appellate courts have occurred, an outcome may be erroneous if it ends up freeing the guilty or convicting the innocent. The fact that a trial has scrupulously followed the letter of the current rules governing the admissibility of evidence and procedures – and thus avoids being slapped down by appellate courts for breaking the rules – is no guarantee of a correct outcome. To the contrary, given that many of the current rules (as we will see in detail in later chapters) are actually conducive to mistaken verdicts, it may well happen that trials that follow the rules are more apt to produce erroneous verdicts than trials that break some of them. Accordingly, our judgment that an error has occurred in a criminal case will have nothing to do with whether the judicial system followed its own rules and everything to do with whether the truly guilty and the truly innocent were correctly identified.
Second, standard discussions of error in the law – even from those authors who, like me, emphasize truth and falsity rather than rule following or rule breaking – tend to define errors only for those cases that reach trial and issue in a verdict. Such authors, naturally enough, distinguish between true and false verdicts. That is surely a legitimate, and an important, distinction, but it is neither the most general nor the most useful way of distinguishing errors. As my definition of “error” has already indicated, I claim that errors occur whenever the innocent are condemned by the system and whenever the guilty fail to be condemned. Obviously, one way in which these mistakes can happen is with a false conviction or a false acquittal. But what are we to say of the guilty person who has been arrested and charged with a crime that he truly committed but against whom charges were subsequently dropped by the prosecutor or dismissed by the judge? These are mistakes just as surely as a false acquittal is. Likewise, if an innocent person – faced with a powerfully inculpatory case – decides to accept a plea bargain and plead guilty, this is an error of the system just as much as a false conviction is, even though the case against the accused is never heard and a jury never renders a verdict.
Clearly, this analysis rests on being able to speak about the truly guilty and the truly innocent. Much nonsense has been creeping of late into several discussions, both popular and academic, of the law. For instance, one often hears it said (in a gross misconstrual of the famous principle of the presumption of innocence) that the accused “is innocent until proven guilty,” as if the pronouncing of the verdict somehow created the facts of the crime. If it were correct that only a guilty verdict or guilty plea could render someone guilty, then there could be no false acquittals, for it would make no sense to say, as the phrase “false acquittal” implies, that a jury acquitted someone who is actually guilty. Since such locutions make perfect sense, we must reject the notion that a verdict somehow creates guilt and innocence.
A second obstacle to talking clearheadedly about guilt and innocence arises from the novel but fashionable tendency to suppose that whether someone is guilty or innocent of a crime simply depends on whether the evidence offered at trial is sufficient to persuade a rational person that the defendant is guilty. The confusion here is more subtle than the former one. It is rooted in the obvious fact that the decision about guilt or innocence made by a reasonable trier of fact will necessarily depend on what he or she comes to learn about the alleged crime. On this view, a verdict is correct so long as it squares with the evidence presented at trial, without making reference to anything that happened in the real world outside the courtroom. One legal scholar, Henry Chambers, has claimed that “what is true is what the [trial] evidence indicates is true.”8 Contrary to Chambers, I claim that nothing that a judge or jury later determines to be the case changes any facts about the crime. Likewise, I claim that, while what is presented in evidence surely shapes the jury’s verdict, that evidence does not define what is true and false about the crime. Unless this were so, it would again make no sense to talk of a true or a false verdict, so long as that verdict represented a reasonable inference from the evidence. Yet, sometimes we come to the conclusion that the evidence presented at trial was deeply unrepresentative of the true facts of the crime. Sometimes, truly innocent people are wrongly convicted and truly guilty people are wrongly acquitted, even though the jury drew the conclusions that were appropriate from the evidence available to them. (Basically, Chambers confuses what I will be calling the validity of a verdict with its truth.)
I will be adamant in insisting that the presumption of innocence, properly understood, does not make a guilty person innocent nor an acquittal of such a person into a nonerror. Likewise, I will argue that verdicts don’t make the facts and neither does the evidence presented at trial; they only give official sanction to a particular hypothesis about those facts. Strictly speaking, the only people innocent are those who did not commit the crime, whatever a jury may conclude about their guilt and regardless of what the available evidence seems to show. Likewise, the truly guilty (those who committed the crime) are guilty even if a jury rationally acquits them. “Being found guilty” and “being guilty” are manifestly not the same thing; neither are “being presumed innocent” and “being innocent.” The naive argument to the effect that what we mean when we say that Jones committed the crime is that a jury would find him guilty utterly confuses questions about what is really the case with questions about judgments issued in the idiosyncratic circumstances that we call criminal trials. There are false acquittals and false convictions, and the existence of each entails that verdicts are not analytically true or self-authenticating. Because they are not, we can speak of verdicts as being erroneous, even when they result from trials that were scrupulously fair, in the sense of being in strict compliance with the rules governing such proceedings. By the same token, we can speak of outcomes or verdicts being true, even when they resulted from trials that made a mockery of the existing rules.
For future reference, it will prove useful to make explicit the moral of this discussion. In brief, it is legitimate, and in some contexts essential, to distinguish between the assertion that “Jones is guilty,” in the sense that he committed the crime, and the assertion that “Jones is guilty,” in the sense that the legal system has condemned him. I propose to call the first sense material guilt (hereinafter, guiltm) and the second probatory guilt (guiltp). Clearly, guiltm does not imply guiltp, nor vice versa.
Similarly, we can distinguish between Jones’s material innocence (innocencem), meaning he did not commit the crime, and his probatory innocence (innocencep), meaning he was acquitted or otherwise released from judicial scrutiny. Again, neither judgment implies the other. With these four simple distinctions in hand, we can combine them in various useful ways. For instance, Jones can be guiltym but innocentp; again, he can be innocentm but guiltyp. Either of these situations would represent an error by the system.
The most basic distinction we need has already been mentioned: that between false inculpatory and false exculpatory findings. These two types of findings are just what one would expect: A false exculpatory finding occurs when the legal system fails to convict a truly guilty felon. A false inculpatory finding is a conviction of an innocent person.
Still, we need to add a couple of other important distinctions to the tool kit of error types. One involves separating valid from invalid verdicts. A verdict of guilty will be valid, as I propose to use that term, provided that the evidence presented at trial establishes, to the relevant standard of proof, that the accused person committed the crime in question. Otherwise, a guilty verdict is invalid. Naturally enough, an acquittal will be valid as long as the conditions for a valid conviction are not satisfied and invalid otherwise. The notion of validity aims to capture something important about the quality of the inferences made by the trier of fact, whether judge or jury. Invalid verdicts can occur in one or both of two ways: a) The trier of fact may give more or less weight to an item of evidence than it genuinely merits, or b) she may misconceive the height of the standard of proof. In either case, the verdict is inferentially flawed.
It is crucial to see that the valid/invalid distinction does not map neatly onto the true/false verdict dichotomy. We settle the truth of a verdict (or what I am calling a finding) by comparing it with the facts. That is, Jones’s conviction is true just in case Jones committed the crime. By contrast, we settle the validity of a verdict by comparing it with the evidence presented at trial, asking whether that evidence meets the applicable standard of proof. Just as a deductive inference can be valid even when its conclusion is false (all horses can fly; all stallions are horses; therefore, all stallions can fly), so a verdict can be simultaneously valid and false. Using the terminology of the previous section, it can be a valid verdict that Jones is guiltyp, even while it is true that Jones is innocentm. By the same token, a verdict of not guilty may be valid even if Jones is guiltym.
Happily, it sometimes turns out that true verdicts are likewise valid ones and that false verdicts are invalid. But neither of these connections is solid. Sometimes, perhaps often, a jury will produce a valid verdict that is false, that is to say, a verdict that reflects an appropriate inference from the evidence presented at trial but that is factually false. This can occur when the evidence admitted at trial, skewed for whatever reasons, invites a conclusion at odds with what actually happened. But even when the evidence is not skewed or unrepresentative of the crime, there is still plenty of scope for a verdict that is valid but not true. Indeed, the standard of proof guarantees as much. Suppose, for the sake of argument, that the standard of proof is something like 95 percent confidence in guilt. A jury hears a case and concludes that it is 80 percent likely that the accused committed the crime. Now, the jury, if it acquits, will be producing a valid verdict, for the rules of proof demand acquittal even when the likelihood of guilt is as high as 80 percent. But that valid verdict is likely to be a false acquittal since, by hypothesis, the likelihood that the defendant committed the crime is quite high.
Likewise, it is easy to conceive how a jury might produce an invalid verdict that was nonetheless true, although these are apt to be less frequent than cases of valid verdicts that are false. What one hopes to achieve, obviously, is a verdict that is both true and valid. We want jurors to convict and acquit the right people and to do so for the right reasons. Both lack of truth and lack of validity will, as I am using the term “error,” represent serious errors of the system, even though they point to quite different ways in which the system has failed. In our efforts to identify the principal sources of error in the legal system, we will be examining rules of evidence and procedure with a view to asking how such rules threaten either the truth or the validity of verdicts.
If the outcome of a criminal proceeding is erroneous in either of these respects – that is to say, if it is either false or invalid (or both) – the system has failed. If one or the other or both types of failure happen frequently, it may be time to change those parts of the system responsible for such errors. In later chapters, we will see that certain practices entrenched in our rules of evidence and procedure tend to produce invalid convictions and acquittals, that is to say, verdicts at odds with what a reasonable person – not bound by those rules – would conclude from the evidence available. Other features of the system, by restricting what can count as legal evidence, tend to produce verdicts that, even if valid, are false. The true/false and valid/invalid distinctions reflect the two primary ways in which a trial verdict may go awry: an inadequate (in the sense of unrepresentative) evidence base or faulty inferences from that base.
There is a third dichotomy that will prove helpful in thinking about sources of error. It distinguishes those erroneous decisions that are reversible from those that are irreversible. For instance, when Schwartz is convicted of a crime, he can appeal the verdict and may persuade a higher court to set that verdict aside. Epistemically, such a review mechanism is invaluable as a way of increasing the likelihood that the final result is correct. By contrast, if Schwartz is acquitted, the verdict cannot be appealed, however flawed may have been the reasoning that led the jury to acquit. Other things being equal, irreversible decisions are more troubling sources of error than reversible ones for the obvious reason that there is no machinery for catching and correcting the former while the latter can, in principle, be discovered and rectified. In due course, we will inquire into the rationale for creating a category of decisions, including verdicts themselves, that is wholly immunized from further review and correction.
Thus far, our focus on error has been principally with the terminal stage, that is, with erroneous verdicts. But many criminal investigations never get as far as this. Sometimes, police investigations simply run out of steam because of lack of clues or bad investigative practices. Although these are errors just as surely as a false acquittal is, they will not be our focus. What will command our attention are those felons who slip through the system, not for lack of incriminating clues known to the police, but who escape trial because of the ways in which the rules of evidence and procedure impede further pursuit of the case against them. These errors will be as revealing a topic of study as false verdicts are.
We need to remind ourselves that a vast number of criminal investigations (probably the overwhelming majority of police inquiries) never reach the trial stage because, although the police have identified a suspect to their own satisfaction, someone or other in authority concludes that the case against him is too weak to take to trial. It may be the police themselves who make this determination or it may be the prosecutor. It can be a grand jury that issues a “no bill,” precluding trial. Or it may be an arraigning judge who dismisses the case. At each of these stages, where a decision must be made whether to proceed along the route to trial or not, the participants are bound by an elaborate body of rules of evidence and procedure. Prosecutors who have in hand a confession know that it may be tossed out if there are doubts about its provenance. Similar questions may arise about much of the other evidence seized by police. Even when prosecutors have powerful evidence of a suspect’s guilt, their decision to proceed to trial must be informed by a calculation on their part as to which parts of the evidence they now have in hand will actually be allowed to go before a jury. If there are rules of admissibility that exclude relevant evidence (and much of this book will address itself to rules of precisely this sort), then those rules will exert a weighty influence not only during the trial itself but on all the preliminary decisions about whether to proceed to trial. Even if we leave aside problems generated by the rules of evidence, the standard of proof likewise works to ensure that many parties who are probably guilty never go to trial. Specifically, prosecutors may believe that the evidence against a suspect strongly suggests that he is guilty but that such evidence would probably be insufficient to persuade a jury of his guilt beyond a reasonable doubt. Short on both financial and human resources, prosecutors are unlikely to proceed with such a case. Judge Richard Posner has put the point succinctly:
Tight [prosecutorial] screening implies that some, perhaps many, guilty people are not prosecuted and that most people who are prosecuted and acquitted are actually guilty.9
It puts the importance of this class of problems into vivid perspective if we remind ourselves that there are far more dismissals than acquittals in the criminal justice system. In federal courts in 1999, for instance, there were about eight judge-ordered dismissals for every acquittal.10 Those writers who focus on the problem of error as if it principally arose in the process of a jury trial itself ignore such numbers at their peril.
This is another way of saying that every year hundreds of thousands of suspects are de facto “acquitted” by prosecutors, judges, and grand juries – without ever going to trial. That is as it should be, since many suspects are surely innocent. Dismissal of charges against an innocent person is not a failure of the system but a success. A failure occurs, in this context, when a guilty suspect has the case against him dropped prior to trial because relevant evidence of his guilt, although in hand, is thought likely not to be admissible if trial were to ensue. Of the three hundred thousand persons suspected of felonies each year – against whom charges are dropped or dismissed before trial – there is every reason to suspect that a certain proportion of these people are guilty. How large that proportion of failures is cannot be ascertained with confidence since the relevant data are inaccessible; instead, our analysis in this book will attempt to determine weak points in the system that may make such false, pretrial “acquittals” much more common than they need be.
A different, more diachronic, way of thinking about the various ways in which failures can occur emerges from imagining a series of filters that mediate between the crime, at the one extreme, and the jury’s verdict, at the other. There is, to begin with, the crime itself. Jones, let us suppose, mugged Smith and stole his wallet. That event is now past. What survive are traces or remnants of the crime. These include memories of the participants and eyewitnesses and physical evidence of the crime (Jones’s fingerprints on Smith’s wallet, contusions on Jones’s face, and so on). The police will come to find some, but rarely all, of these traces. If they and the prosecutor decide that they have a solid case against Jones, they will next have to persuade a judge or a grand jury that the case is strong enough to go forward. Supposing that all these hurdles have been leapt, the prosecutor will now choose from among the traces known to the police a subset that he intends to enter as evidence at the trial. Jones’s attorney will make a similar decision. At Jones’s pretrial evidentiary hearing, a judge will decide which of these submitted traces can be revealed to the jury. Once the evidence questions are settled, the judge may wrongly decide to dismiss the charges against the accused. Once the trial begins, if it gets that far, the now heavily filtered evidence will be presented and subjected to cross-examination. Once both sides have had their say, the judge will instruct the jury about the relevant law that Jones is alleged to have broken and on the threshold of proof that they should use in deciding whether to convict Jones.
There is, obviously, ample scope for error at each of these stages. Important evidence, either inculpatory or exculpatory, may elude the best efforts of prosecution and defense to find it. The prosecution may find exculpatory evidence but suppress it, and the defense may be aware of inculpatory evidence that it “forgets” to mention. The judge’s rulings on the admissibility of evidence submitted for trial may end up including evidence that is likely to mislead the jury or excluding evidence that the jury should hear. The grand jury may err in their decision to indict. The defendant may refuse to testify or subpoenaed witnesses with important information may disappear. Witnesses with relevant evidence might not be called because both prosecution and defense fear that their testimony would undermine their respective cases. The judge may misinstruct the jury with respect to the relevant law or botch the instructions about the standard of proof – which occurs more often than you might imagine. (For details, see the next chapter.) Even if all else goes properly, the jury may draw inappropriate inferences about guilt or innocence from the evidence before them or they may misunderstand the level of proof required for a conviction. Even once the verdict is pronounced, the room for error has not disappeared. If the jury voted to convict, the accused may file an appeal. Appellate courts may refuse to hear it, even when the defendant is innocent. Or, they may take the appeal but reverse it when the verdict is sound or endorse the verdict when it is false. If the defendant is acquitted, double jeopardy precludes further appeal, even if the trial was riddled with acquittal-enhancing errors.
Eliminating all these possible sources of error (and I have mentioned only the more obvious) is clearly impossible. The aim of the justice system, realistically construed, should be to attempt to reduce them as far as possible. Current evidential practice in the criminal law, as we will see, often fails to do that. Worse, it frequently increases the likelihood of error deliberately by adopting rules and procedures that prevent the jury from learning highly important things about the crime.
The charge that I have just made can be put in slightly more technical terms, and it will probably be useful to do so. In all reasoning about human affairs (and other contingent events), there are two key concepts regarding evidence that must be grasped. One is called credibility or sometimes (as in the law) reliability. As the term suggests, a piece of evidence or testimony is credible when there is reason to believe it to be true or at least plausible. The other pertinent concept is known, in both the law and in common sense, as evidential relevance. The core idea is that a piece of information is relevant to the evaluation of a hypothesis just in case, if credible, it makes that hypothesis more or less probable than it was before. If a certain bit of information, even when credible, would not alter our confidence in a hypothesis one way or the other, we deem it irrelevant to that hypothesis. In the criminal law, there are always two key hypotheses in play: a) A crime was committed and b) the defendant committed it. Any testimony or physical evidence that would make a reasonable person either more inclined or less inclined to accept either of these hypotheses is relevant. Everything else is irrelevant.
Both credibility and relevance are crucial to qualify something as germane evidence. Jurors, above all others, must assess both the credibility and the relevance of the evidence they see and hear. For reasons having roots very deep in the common law, however, the judge in a criminal trial is generally not supposed to let judgments of credibility enter into his or her decision about the acceptability of proffered evidence. This is because the jury, rather than the judge, is by tradition charged with determining the “facts” of the case. Deciding whether eyewitness testimony or physical evidence is credible would, in effect, imply a decision about its facticity. Since that is the province of the jury rather than the judge, the usual pattern is for judges to rule on relevance but not on reliability. This means that when judges make decisions about relevance, they are obliged to think hypothetically; that is, they must ask themselves, “if this evidence were credible, would it have a bearing on the case?” This is why, when a judge admits evidence as relevant, nothing is implied with respect to its credibility. (A significant exception to this principle occurs in decisions about the admission of expert testimony, where the judge is specifically charged to determine whether the basis for the testimony of the avowed expert is “reliable.”11)
American courts at every level of jurisdiction accept this notion of relevance. One of the important and legitimate gate-keeping functions of a judge is to see to it that the jury hears all and only relevant evidence. If American judges stuck resolutely to this principle, they could not be faulted on epistemic grounds since virtually all forms of sophisticated hypothesis evaluation (in science, medicine, and technology, for instance) work with this same notion of relevance.
Unfortunately, however, legal texts and the practices of courts routinely flout the relevance-only principle. This is because judges have a second criterion they use, alongside the demand for relevant evidence. It is often known as the admissibility requirement. To be admissible, evidence must not only be relevant; it must also meet a variety of other demands. For instance, the evidence cannot have been acquired by a violation of the rights of the accused. The evidence cannot arise from privileged relations that the accused has had with various professionals or his spouse. The evidence generally cannot have been obtained illegally, even if its being seized violated none of the rights of the accused. The evidence cannot be such that it might inflame the passions of the jurors or unfairly cast the defendant in an unfavorable light. The evidence cannot inform the jury that the defendant withdrew a confession of guilt, nor can it refer to admissions of guilt made by the defendant during negotiations about copping a plea. The evidence generally cannot come from a witness whose testimony would be self-incriminating. The jury cannot be informed when key witnesses escaped giving testimony by claiming their Fifth Amendment rights. The jury cannot be told whether the accused cooperated with the police in their inquiries. If the accused does not offer testimony on his own behalf, the judge explicitly instructs the jury to ignore that relevant fact, rather than supposing that the accused may have something to hide.
Virtually no one disputes that information of all these sorts is relevant in the technical sense, for it indubitably bears on the probability of the hypothesis that the defendant is guilty. In most jurisdictions, however, these and many other examples of admittedly relevant evidence will not be admitted during the trial. Subsequent chapters will describe many of these exclusionary principles in detail. What we should note here is that every rule that leads to the exclusion of relevant evidence is epistemically suspect.12
It is universally agreed, outside the law courts, that decision makers can make the best and most informed decisions only if they are made aware of as much relevant evidence as possible. Excluding relevant but nonredundant evidence, for whatever reasons, decreases the likelihood that rational decision makers will reach a correct conclusion. Accordingly, we will want to examine these exclusionary principles carefully to see whether the damage they inflict on our truthseeking interests are suitably balanced by gains of other sorts.
It might be instructive to include here one example of this distinction between relevance and admissibility in order to put some flesh on the skeleton of abstractions with which we have been working. A paradigmatic example of the problems we will be facing throughout the rest of the book is provided by the law’s unimpressive efforts to distinguish between evidence that is “unfairly prejudicial” and evidence that is not.
At the preliminary hearing preceding a trial, both sides describe the evidence they intend to present at trial and argue about its admissibility. Despite the rule to the effect that the judge should generally admit relevant evidence, the law gives her enormous discretion to exclude evidence, however relevant and however inculpatory, if in her judgment that evidence is of such a sensational or inflammatory nature that ordinary jurors would be unable to assign it its true weight. Specifically, the judge is supposed to conduct a balancing test that ultimately comes down to this question: Is the probative power of this evidence sufficient to offset its prejudicial effects in warping the judgment of jurors? If the answer to that question is affirmative, it should be admitted; otherwise, by law it is to be excluded. To be precise, federal evidence law says:
Although relevant, evidence may be excluded if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of the issues, or misleading the jury, or by considerations of undue delay, waste of time, or needless presentation of cumulative evidence.13
Key here is the notion of “unfair prejudice.” There are a great many things that courts have held to be apt to prejudice a jury unfairly. They include evidence that the defendant has a bad or violent character, especially vivid and gruesome depictions of the crime, and evidence of the defendant’s association with causes or persons likely to evoke hostility or antipathy from jurors. The same doctrine has been used to justify excluding the confession of a nontestifying codefendant that mentions the defendant’s participation in a crime,14 graphic photos of the corpse of a homicide victim,15 and samples of bloodstained clothing of the victim of an assault.16
The problem, of course, is that information of this kind is often powerful evidence of the defendant’s guilt. Excluding it can weaken the case against the defendant substantially while, if it is really prejudicial, admitting it makes it more likely that jurors may make their decision on purely visceral grounds. Put in slightly more technical language, the judge is required to make a ruling about evidence that, if admitted, may lead to a false conviction while, if suppressed, may lead to a false acquittal. As we have seen, the judge is supposed to balance these two concerns against one another and decide about admissibility accordingly.
It may help to describe the problem a little more abstractly. In cases of this sort, the judge is called on to decide which of two quantities is greater: the probability of inferential error by the jury if the contested evidence is admitted (which I shall symbolize as “prob [error with e]”) versus the probability of inferential error if the contested evidence is excluded (prob [error excluding e]). The first sort of error represents a potential false conviction; the second, a potential false acquittal. In making her decision about admitting or excluding e, the judge must perform an incredibly difficult task: She must decide on the relative likelihood of the two errors that may arise – that is, she must assign rough-and-ready values to prob (error with e) and to prob (error excluding e).
It seems doubtful whether this decision can be made objectively. To decide on the values of prob (error with e) and prob (error excluding e), a judge needs much more data than we currently have in hand about the likelihood that particular pieces of evidence (such as vivid, gory photos of the crime scene) will distort a jury’s ability to give such evidence its legitimate weight. Well-designed empirical studies on the prejudicial effects of various sorts of evidence are extremely scarce. Even worse, collecting such information would be inherently difficult since researchers would have to be able to distinguish the emotional impact of a bit of evidence from its rational probative weight. No one has proposed a design for an empirical test subtle enough to make that distinction.
I do not mean to convey the impression that this decision about admitting relevant but potentially inflammatory evidence is always insoluble. Sometimes, the problem admits of an easy solution. For instance, the prosecution may have other types of evidence, apparently less unfairly prejudicial, that will permit the state to make its point, in which case the exclusion is no big deal (since the prejudicial evidence here is clearly redundant, and redundancy is always a legitimate ground for exclusion). But what is a judge to do when a principal part of the prosecution’s case involves evidence that, while highly inculpatory, may also appear “unfairly prejudicial” and where no other evidence will do?
Consider a hypothetical example: Smith is charged with being a member of a gang that entered a busy restaurant at midday, tossing grenades, firing weapons, and generally creating mayhem. By chance, one patron of the restaurant took photographs during the assault, before he was himself gunned down. One photo in particular is at issue. It shows Smith lobbing a grenade into one corner of the restaurant and also shows, in vivid color, mangled body parts and blood galore and is generally a horribly graphic depiction of the crime scene. The photo obviously passes the relevancy test. It apparently depicts the accused committing the crime with which he is charged. It is not merely relevant but highly relevant. If we suppose that no witnesses survived the mayhem, it is uniquely powerful in placing Smith at the center of things.
Unfortunately, however, the judge also considers the photograph to be so vivid and awful that it invites a purely visceral reaction from the jurors. Seeing blood and gore depicted in this manner may, she fears, incline the jurors to rush to judgment rather than considering objectively the other evidence in the case, some of which may be exculpatory. Without the photo, the jury may well acquit Smith since there were no eyewitnesses. With the photo, reckons the judge, they will surely convict. Should the judge admit the photograph into evidence? Currently, that decision is left entirely up to her, with precious little assistance from the law. The guiding legal principle, as we have seen, is that the evidence should be excluded if it is more “unfairly prejudicial” than it is probative. Curiously, the law of evidence includes no canonical definition of when a sample of evidence is “unfairly prejudicial,” apart from this gem of unclarity in Rule 403: “‘Unfair prejudice’ within its context means an undue tendency to suggest decision on an improper basis, commonly, though not necessarily, an emotional one.” Like pornography, unfair prejudice seems to be the sort of thing that, while it eludes definition, one can recognize when one sees it. But this won’t do. As Victor Gold has noted: “Absent a coherent theory of unfair prejudice, trial courts cannot meaningfully evaluate evidence on or off the record for the presence of unfair prejudice, nor can they conduct the required balancing test.”17 How, in such circumstances, is a judge supposed to do this “balancing” to decide whether “its probative value is substantially outweighed by the danger of unfair prejudice”?
One might argue that this particular rule of evidence is not really so offensive epistemologically since in practice it should lead only to the exclusion of inflammatory evidence that is relatively nonprobative. After all, the rule itself seems to concede that, if the evidence is of very high probative value, it could be excluded only in those circumstances where its unfairly prejudicial character was even greater than its probativeness. But there are plenty of actual cases that give one some pause as to how often the weight of highly relevant evidence really is allowed to trump its being even mildly prejudicial.
Consider two real examples of the kind of balancing that goes on when trial and appellate judges try to assess unfair prejudice. In a 1994 case in south Texas, Ramón Garcia was accused of burgling Charles Webster’s house. Garcia was seen in the house at the time of the burglary by a police officer who had been called to the scene by a neighbor. Taking flight, Garcia was subsequently caught. Police found no contraband on Garcia himself when he was apprehended, but several items stolen from Webster were found on the ground near the site of his arrest. By way of showing intent to commit burglary, the prosecutor introduced evidence that Garcia had arrived at the scene of the crime on a bicycle that he had stolen from a neighboring house two days earlier. The boy whose bicycle was stolen by Garcia testified that he was its owner. Garcia was convicted. His attorney appealed, arguing that the evidence of the stolen bicycle unfairly prejudiced the jury against his client. The appellate court, siding with Garcia, did not deny that the evidence of the stolen bicycle was relevant to the question of whether Garcia intended to rob the Websters but held that its relevance was outweighed by its unfairly prejudicial nature, “particularly so when the State chose to offer the evidence through the child rather than his parents.”18
The logic of the appellate ruling is a bit tortuous, but here is what seems to be going on: Besides conceding the relevance of the fact that the defendant arrived at the scene of a burglary on a stolen bicycle to the question of Garcia’s intention to rob the Websters, the superior court even seems to grant that evidence concerning the theft of the bicycle might not have been unfairly prejudicial if the testimony about its theft had been offered by an adult. But for a child to testify that his bicycle had been stolen seems to have the court in a tizzy for fear, I suppose, that the jury will conclude that anyone who would steal a bicycle from a young boy must be very bad indeed and deserves to be sent to jail, whether he robbed the Websters or not. Are we then to conclude that whenever children have inculpatory evidence to offer, the specter of unfair prejudice is raised and that their evidence should be excluded? I doubt that is the intended moral, but the example certainly suggests how subjective the determination of unfair prejudice can sometimes be.
Consider briefly a second case, also from Texas, where the balancing test seems to have gone awry. In 1992, Kenneth Nolen was accused of aggravated possession of methamphetamine. Acting on a warrant, police found Nolen asleep in the bedroom of a friend’s house. Smelling a strong odor, they opened the bathroom door, next to which Nolan was sleeping, and discovered a laboratory for making amphetamine. Nolen’s prints were found on the lab equipment. To convict someone in Texas of aggravated possession, the state must show that the accused was aware of the fact that the drug in his possession was an illegal substance. Nolen’s attorney suggested that, although his client had indeed been making amphetamines, he did not know that such activity was illegal. To counter that suggestion, the prosecutor sought to introduce evidence that Nolen had been convicted three years earlier of breaking into the evidence room of the local sheriff’s office to steal laboratory equipment suitable for making amphetamines. The prosecutor argued, and the trial judge agreed, that the earlier theft of such equipment was highly relevant to the question whether Nolen knew enough about amphetamines to realize that they were illegal. In the prosecutor’s closing arguments, he insisted that “it’s a reasonable deduction from the evidence that [if] that man is so daring [as] to take lab equipment from the Hood County Sheriff, he certainly knows about am[p]hetamine and the equipment used to produce amphetamine.”19
The jury convicted Nolen. On appeal, a higher court reversed the verdict, insisting that “it was an abuse of [the judge’s] discretion to determine that the extraneous evidence of burglarizing the sheriffs evidence shelter was not substantially outweighed by the danger of unfair prejudice to Nolen, confusion of the issues, and misleading the jury.”20 How precisely is it unfairly prejudicial to Nolen – facing a charge of knowingly making illicit drugs and with his prints all over the equipment in question – to show that he had previously stolen such equipment from the sheriff’s office? Since what was in dispute was whether Nolen knew that making methamphetamine was illegal, we would seem to have here evidence highly germane to the hypothesis that Nolen was knowledgeable about drug making. Not so, says the appellate court, since it is not “deductively certain” that “a man who steals glassware certainly knows the characteristics of a particular chemical compound that may be produced with that type of glassware.”21 Well, yes. It is (just about) conceivable that Nolen stole equipment for making amphetamines from the evidence room of the sheriffs department without knowing what such equipment was used for and without knowing that making such stuff was illegal. But we should not be looking for deductive truths in the law. The question is whether Nolen’s previous conviction for theft of equipment for making amphetamines has a powerful evidential bearing on the question of whether he knew three years later, while making amphetamines, that what he was doing was against the law. For a court to hold that such evidence is more likely to be unfairly prejudicial than relevant strikes me as extraordinarily obtuse. (I am not claiming that cases such as these two are the norm, but, at a minimum, they suggest that the balancing test demanded by the unfair prejudice rule is, at best, problematic.)
Surely, a preferable alternative would be to admit all evidence that is genuinely relevant, accompanied, where appropriate, by an explicit reminder from judge to jury to bring their critical faculties to bear in evaluating the relation of the evidence to the crime and in keeping their emotional reactions to the evidence firmly in check. Of course, we do not know how earnestly juries could or would follow such an instruction. But, for that matter, neither do we really know which kinds of evidence unfairly warp jurors’ judgment and which do not. Since the judge has no robust empirical information about the latter issue, her decision about which potentially prejudicial evidence to include and which to exclude is likely to be every bit as suspect as an emotionally driven verdict from the jury.
It is not only the judge who has a role to play here in encouraging the jury to stay on the straight and narrow. One of the functions of the adversarial system is to give each side a shot at undermining or otherwise calling into question the case presented by the other. If there is evidence that the defense regards as misleading or that it suspects may otherwise steer a jury in the wrong direction, it is the job of defense counsel to seek to fit that evidence into a context favorable to the defendant, if that is possible. Failing that, it falls to defense counsel to persuade the jury not to attach more weight to any specimen of inculpatory evidence than it duly deserves. Like the judge, counsel may fail in this task from time to time. Jurors may conceivably rush to judgment for all sorts of inappropriate reasons, despite having been warned of the dangers of doing so.
That conceded, if we cannot generally trust jurors to keep their emotions in check, then we should abandon trial by jury altogether. The very idea of trial by jury depends on the presumed fairness and common sense of twelve peers of the accused. If jurors cannot generally give vivid but relevant evidence its appropriate weight – having been warned by the judge to do so and having heard counsel for each side make its pitch about the meaning of the evidence – then the system is rotten to the core. Paternalistically coddling jurors by shielding them from evidence that some judge intuits to be beyond their powers to reason about coherently is not a promising recipe for finding out the truth.
I use the term “paternalism” deliberately. Recall that, in a bench trial, the same judge who decides on the admissibility of evidence usually acts as the trier of fact. In short, the system trusts judges to be able to see inflammatory and unfairly prejudicial evidence and then to be able to put it into perspective, not allowing it to warp their judgment. By permitting judges in bench trials to see evidence that we would not permit juries to see, we are saying that juries are less reasonable, less objective, or less mature than judges. That may be so; as far as I know, the issue is unsettled. But, settled or not, this is not what judges are for in an adversarial system. Their job, apart from generally maintaining order in the court, is to explain to jurors what the law means. That is what they are trained to do, and there is nothing paternal about that role. It becomes paternal when, out of a systemic distrust of the good sense of jurors, we cast the judge in the role of arbiter on both empirical and policy questions that should not be hers to settle.
Put in grander terms, it will be the recurring theme of this book that, leaving redundancy aside, the only factor that should determine the admissibility or inadmissibility of a bit of evidence is its relevance to the hypothesis that a crime occurred and that the defendant committed it. The exclusion of admittedly relevant evidence on the grounds of its unfairly prejudicial character is motivated by commendable epistemic instincts. But the rule itself requires judges both to have empirical knowledge that they lack and to make policy determinations (for instance, about the relative seriousness of false acquittals and false convictions) that are beyond their ken.
As we have seen, my proposal is squarely at odds with existing practice. In American courts, “mere” relevance, even powerful relevance, does not ensure admissibility. As the U.S. Supreme Court argued in a famous case, Michelson v. U.S.:
The State may not show defendant’s prior trouble with the law, specific criminal acts, or ill name among his neighbors, even though such facts might logically be persuasive that he is by propensity a probable perpetrator of the crime. The inquiry is not rejected because character is irrelevant; on the contrary, it is said to weigh too much with the jury and to so overpersuade them as to prejudge one with a bad general record and deny him a fair opportunity to defend against a particular charge. The overriding policy of excluding such evidence, despite its admitted probative value, is the practical experience that its disallowance tends to prevent confusion of issues, unfair surprise and undue prejudice.22
This token of conventional folk wisdom may be grounded in the “practical experience” of the judiciary. But precious few well-designed empirical studies bear out the claim that properly instructed jurors, exposed to the confrontations typical of the adversary system, are incapable of giving inflammatory or prejudicial but relevant evidence the weight it rationally deserves. Since that is so, rules of admissibility that trump relevance cannot be shown to further epistemic ends, not even when those rules (such as the one against unfairly prejudicial evidence) are couched in epistemic terms (“preventing a jury’s verdict from being shaped by prejudice rather than the facts”). On the contrary, they almost invariably thwart those ends by keeping obviously relevant evidence out of the courtroom. The proof of that thesis is the thrust of much of the rest of this book.
Notes
1 Criminal Law Revision Committee, Eleventh Report, Evidence (General) 1972, Cmnd. 4991. at §§62–4.
2 From Powell’s dissent in Bullington v. Missouri, 451 U.S. 430 (1981).
3 Tehan v. U.S., 383 U.S. 406, at 416 (1966).
4 Lest you take my remarks about the lack of a coherent design in the rules of trials as casting aspersions on the founding fathers, I hasten to add that the system now in place is one that they would scarcely recognize, if they recognized it at all. Many of the features of American criminal justice that work against the interests of finding truth and avoiding error-features that we will discuss in detail later on – were additions, supplements, or sometimes patent transformations of American criminal practice as it existed at the beginning of the nineteenth century. Congress or state legislatures imposed some of these changes; judges themselves created the vast majority as remedies for serious problems posed by the common law or abusive police practices. A few date from the late-nineteenth century; most, from the twentieth.
5 To see the point of the scare quotes, consult Chapter 7, where we will observe that the majority of “coerced” confessions are not coerced in the lay sense of that term.
6 Accordingly, I ask those readers who know the fine points of the practice of the law far better than I do to overlook the occasional acts of ignorance on my part, of which there are doubtless several, unless they actually impinge upon the cogency of the argument that I am making.
7 Courts typically distinguish between errors that, while acknowledged as errors, did not decisively affect the outcome of a trial (called “harmless errors”) and more serious errors, which call for retrial or reversal of a conviction.
8 Henry Chambers, Reasonable Certainty and Reasonable Doubt, 81 MARQ. L. REV. 655, at 668 (1998).
9 Richard Posner, An Economic Approach to the Law of Evidence, 51 STAN. L. REV. 1477, at 1506 (1999).
10 See the Department of Justice, SOURCEBOOK OF CRIMINAL JUSTICE STATISTICS FOR 1999, Table 5.16. For a thorough discussion of this issue, see Samuel R. Gross, The Risks of Death: Why Erroneous Convictions Are Common in Capital Cases, 44 Buffalo L. REV. 469 (1996).
11 The Supreme Court has held that “the trial judge must ensure that any and all scientific testimony or evidence admitted is not only relevant, but reliable” (Daubert v. Merrell Dow Pharms., 113 S. Ct. 2786, at 2795 [U.S. 1993]).
12 The only time when it is obviously appropriate to exclude relevant evidence is when it is redundant with respect to evidence already admitted. Testimony from two hundred witnesses asserting X is scarcely better than that from two or three credible ones asserting X, unless X happens to be a very bizarre event.
13 Federal Rules of Evidence, Rule 403.
14 Bruton v. U.S., 391 U.S. 123 (1968).
15 State v. Lafferty, 749 P.2d 1239 (Utah 1988).
16 State v. White, 880 P.2d 18 (Utah 1994).
17 Victor Gold, Observations on the Nature of Unfairly Prejudicial Evidence, 58 WASH. L. REV. 497, at 502 (1983).
18 Garcia v. State, 893 S.W.2d 17, at 22 (Tex. App. 1994).
19 Nolen v. State, 872 S.W.2d 807, at 813 (Tex. App. 1994).
20 Ibid., at 814.
21 Ibid., at 813.
22 Michelson v. U.S., 335 U.S., at 475–6 (1948).