Often when we obey a speed limit we are driving at a speed that is not the same as what we think is the best speed given the traffic, the driving conditions, and our own driving skills. Consequently, to obey a speed limit is to do something we do not think best. Similarly, making a decision just because the same decision has been made before-following precedent-gets interesting primarily when we would otherwise have made a different decision. The parent who gives the younger child the same privileges at the same age as an older child feels the pull of precedent only when he or she otherwise thinks there is a good reason for treating the two differently, and so being constrained by precedent is again a path away from what had otherwise seemed to be the right decision. And we say we are obeying or following an authority only if what we are doing because of what the authority has said is not the same as what we would have done if left to our own devices to make the decision we thought best. The soldier who follows an order might well do something else if allowed to make his or her own unguided (or uncommanded) decision, just as the obedient student or child is one who suppresses his or her own desires to do something else.

It is important to understand that the belief that there is a moderately distinct form of reasoning we can call "legal reasoning" is in the final analysis an empirical claim. Most people can describe a unicorn, but our ability to describe a unicorn is not inconsistent with the crucial fact that there are no actual unicorns in the world. Similarly, most of us can point to examples of genuine self-sacrificing heroism, but even as we do so we recognize that such behavior is highly unusual. And so the lesson we should draw from these examples is that our ability to describe legal reasoning, and even to point to actual examples of its use, says less than is commonly supposed about how often such reasoning is an important component of what lawyers and judges actually do. To point to one or a few instances of genuine constraint by precedent, for example, says almost nothing about the frequency of such constraint throughout the universe of legal decisions.

In theory, it would often be possible for the rule simply to be a restatement of the background justification. A few years ago, for example, the state of Montana eliminated all fixed speed limits, requiring instead only that driving should be "reasonable and prudent."' But drivers have widely divergent ideas of what is reasonable and what is prudent, and so do police officers and judges. As a result, there developed wide variations in speed limit enforcement, the consequence being that drivers became highly uncertain about just how fast they could go without running afoul of the law. This much uncertainty was too much for the Montana Supreme Court, which struck down the "reasonable and prudent" rule as excessively vague.

Because genuinely easy cases and straightforward applications of legal rules are so rarely disputed in court, the array of disputes that do wind up in court represents a skewed sample of legal events. The effect, known as the selection effect," is such that the cases that wind up in court are only-or almost only-the ones in which two opposing parties holding mutually exclusive views about some legal question both believe they have a reasonable chance of winning.

The selection effect is the major factor in determining which disputes or law-controlled events wind up in litigation, but the effect is even greater as we proceed up the appellate ladder. In its 2007 Term, for example, the United States Supreme Court, which has almost total power to decide which cases it wants to hear, was asked to hear more than nine thousand cases from the federal courts of appeals and from the highest courts of the states, but agreed to take and decide, with full briefing and argument and opinions, only seventy-one. These seventy-one cases were almost all ones in which there was no clear legal answer, and taking these seventy-one as representative of how law works or how rules work would be a major blunder.

For example, the Seventh Amendment to the Constitution provides the right to a jury trial in any civil case at common law in a federal court in which the amount in controversy is "twenty dollars," and it is obvious that the purpose behind the twenty-dollar minimum was to limit jury trials to cases in which substantial sums were involved. But although twenty dollars was a substantial amount of money in 1791, when the Seventh Amendment was adopted, it is hardly substantial anymore.

The rule includes or encompasses instances that the background justification behind the rule would not cover, as in the Kirby and Church of the Holy Trinity cases, as with the driver driving safely at 70, and as with an ambulance which might fall within the literal scope of the "no vehicles in the park" rule. In such cases the reach of a rule is broader than the reach of its background justification, and so we say that the rule is overinclusive.

Justice Peckham's formalism was the formalism of disingenuousness and fully entitled to the stigma it has attracted. When we look at another conception of formalism, however, the formalism of Justice Thurgood Marshall in United States v. Locke, for example, it is not so clear that formalism deserves to be treated as a vice at all. It is, to be sure, formalistic to take the literal meaning of the words "prior to December 31" in United States v. Locke as dictating a result other than what seems to be the most sensible one, because it is to treat the form of a legal rule as more important than its deeper purpose, or more important than reaching the best all-things-considered judgment in the particular context of a particular case. But although Locke is from this perspective formalistic, it is also formalistic in just the same way to use the 55-mile-per-hour speed limit to penalize the driver who is driving safely at 70, to penalize the short-swing trader who in fact has no inside information, to allow those with twenty-one-dollar claims to demand a jury trial, and to prohibit otherwise qualified thirty-four-year-olds from becoming president. In all of these cases, law operates formally in treating the meaning of the words of a rule as more important than achieving the law's deeper purpose and reaching the ideal result in this particular case. Formalist this may be, but formalism is, as these and countless other examples demonstrate, a central feature of what makes law distinctive.

The importance of what a rule actually says is not just a point about rules. More pervasively, to ignore the even more ubiquitous importance of what rules do even when what they do appears unfair is to ignore something very important about law itself. It is not law's purpose, of course, to be unfair for the sake of being unfair. But there is an important group of values-predictability of result, uniformity of treatment (treating like cases alike), and fear of granting unfettered discretion to individual decision-makers even if they happen to be wearing black robes-that the legal system, especially, thinks it valuable to preserve. These values often go by the name of the Rule of Law, and many of the virtues of the Rule of Law are ones that are accomplished by taking rules seriously as rules. In doing so, law remains irreducibly formal and thus at times seemingly unfair in particular cases. But law is more than simply doing the right thing in each individual case. At times law's unwillingness to do just that will seem wrong, but what makes law what it is-usually for better but sometimes for worse-is that it takes larger institutional and systemic values as important, even if occasionally at the expense of justice or wise policy or efficiency in the individual case.

With respect to the former, which is not really precedential reasoning at all, the instant court may learn from a previous case, or be persuaded by some decision in the past, but the decision to do what another court has done on an earlier occasion is not based on the previous case's status as a precedent. Instead the decision exemplifies the fundamental human capacity to learn from others and from the past. There are many instances in which the instant court will be persuaded by the reasoning of another court, but if the instant court is genuinely persuaded, then it is not relying on-obeyingprecedent at all.' To see why this is so, consider a simple nonlegal example: Suppose I am boiling an egg. I boil it for six minutes, and am pleased to discover that it is cooked to precisely my preferred hardness. Consequently, the next time I boil an egg I do so, not surprisingly, for six minutes. I have learned from the previous "case," but when I boil the second egg for six minutes, I am not boiling it for six minutes because I boiled it for six minutes on the previous occasion. I am boiling it for six minutes because six minutes is the right time. I know this because I have learned from the previous action, but on subsequent occasions I make the decision because of what I then know.

In contrast, reasoning from precedent-and maybe it is a mistake to call it "reasoning" at all-is following a previous decision just because of its status as a decision of a higher court or of the same court on an earlier occasion, not because the follower in the instant case has been persuaded by the reasoning of the precedent case. Some lower court judge in New Jersey, for example, might still believe after Henningsen that all nonfraudulent contractual provisions should be strictly enforced according to their terms and that the Henningsen court's concern for the consumer was misplaced. Even after reading Henningsen, he remains unpersuaded. Yet however much he continues to believe in the strict enforceability of written provisions, and even though he believes Henningsen to have been wrongly decided, he is still obliged, as a lower court judge in the same jurisdiction, to follow Henningsen despite being convinced of its error. So too for stare decisis. If in 1970, ten years after Henningsen, the majority of the New Jersey Supreme Court consisted of justices not on the court at the time of Henningsen, and if those new justices believed Henningsen to have been erroneously decided, the obligations of stare decisis would still have obliged them to decide the same issue in the same way. They would have been constrained to follow a decision they thought mistaken just because of its existence as a previous decision of the same court. The British legal theorist P. S. Atiyah puts it directly: "The concept of a system of precedent is that it constrains judges in some cases to follow decisions they do not agree with."

When we turn to horizontal precedent, however, the arguments in its favor are less obvious. Stare decisis is a pervasive principle of the common law,' but it is far less so in nonlegal contexts. Scientists, for example, are not expected to reach the same conclusions as their predecessors just because their predecessors have reached them. It would be surprising if Congress were to make the same decisions as previous Congresses only because previous Congresses had made them. And no one believes that presidents should follow those decisions of their predecessors with which they disagree. Indeed, we often elect them not to. Thus it is no surprise that books about logic typically treat arguments from precedent as fallacies, because the fact that someone has reached a conclusion in the past says nothing about whether it is the correct conclusion now.

If in such cases the bare statement of the facts and the outcome cannot tell us what the precedent case fully "stands for," then it is tempting to say that the question of legal similarity is itself determined by the law. That is why discussions of precedent, including Goodhart's, commonly talk not about facts but about material facts. In concluding that the holding of the case-the term more common in the United States than ratio decidendi, although there are slight differences in meaning-is a combination of the material facts and the outcome, Goodhart and others solved the level of generality problem, but at the cost of undermining the core of their view. So to Goodhart it would have been an error to say that the car being a Buick was a material fact, because the car's "Buickness" was no more material than the fact that Mr. MacPherson's last name began with the letter "M." Under this view, a fact is material when a legal rule makes it legally important. It is a legal rule that tells us when two things are similar, and thus it is a legal rule that tells us the level of generality at which the facts should be understood and described by the deciding court. It is therefore a legal rule that would tell us that "automobile" is a legally material category while "Buick" is not. But although this is often so, relying on a legal rule to tell us which cases are materially similar and which are not avoids the very question we are trying to answer. If the standard for materiality comes from outside the precedent case-a statute, for example-then the statute is doing the work and we do not have an example of precedential constraint at all.

So in Marbury v. Chief Justice John Marshall held that the judiciary Act of 1789, upon which the subject-matter jurisdiction of the Court had been asserted, was unconstitutional. But he also went on to say that the Supreme Court possessed the power to exercise jurisdiction over the president of the United States, a conclusion that infuriated President Thomas Jefferson, not least because it was wholly unnecessary to the Court's conclusion and thus clearly dicta. If the Court had no subject matter jurisdiction after all, then there was no need for it to say anything at all about who would have been subject to that hypothetical jurisdiction.

Yet if providing reasons for their decisions is part of what we expect courts to do, and if providing reasons is a key to the actual workability of a system of precedent, then the traditional distinction between holding and dicta may be more problematic than commonly thought. Because a reason is necessarily broader than the outcome that it is a reason giving a reason is saying something broader then necessary to decide the particular case. And that seems to be dicta. What is technically dictanot totally necessary for the result-is precisely what it is that makes it possible for us to generalize from a very specific ruling and thus to use it as a precedent in the future.

Not so, however, when an advocate or a lower court judge can distinguish the instant case from the precedent case. Although it might be said that a binding precedent from a higher court simply obliges the lower court to follow it, it would be more accurate to say that a binding precedent obliges a lower court to follow it or to distinguish it from the instant case. In practice, a great deal of legal argument involves the attempt by one side to claim that some higher court case controls the result in the instant case, while the other side insists that there is a sufficient distinction between the two that the outcome in the precedent case need not be the outcome in the instant case.

Although courts may occasionally overrule their own previous decisions, doing so requires more than just the belief that the previous de cision was in error. If that were all that were necessary, stare decisis would become meaningless, because it is precisely the point of stare decisis that a court should treat a previous decision as binding just because of its existence and not because it is perceived to be correct. If every time a court believed an earlier decision to be mistaken it could overrule that decision, then there would be no principle of stare decisis at all.

It is not that the government may never, under the equal protection clause of the Fourteenth Amendment, draw a distinction based on race. Rather, the government may draw such a distinction only if it has a compelling interest in doing so, and the fact that the interest must be compelling (rather than simply substantial, legitimate, or rational) shows exactly how a right, like a rule and like an obligation, can make a genuine difference without being absolute. What there is a reason to do is different from what should be done, all things considered, just as what there is a right to do is different from what the right-holder actually gets to do, all things considered.

It was formerly the practice in English courts, for example, to treat as impermissible in a lawyer's argument or a judge's opinion a reference to a secondary source written by a still-living author. If the author of a treatise or (rarely) an article was dead, then citation was permissible, but not otherwise. The reasons for this practice remain somewhat obscure, although apparently it developed out of a concern that it was far easier for those who are still living than those who are dead to change their minds. What is important, however, is that the prohibition gradually withered, a withering that commenced more or less with the citation by the House of Lords in 1945 to a work by the then still-living Arthur Goodhart. Once the first citation to a living secondary author appeared, subsequent courts became slightly less hesitant to do the same thing, and then less hesitant yet, and over time the practice became more widely acceptable.

A citation to a particular source is not only a statement by the one citing it that this is a good source, but is also a statement by the citer (especially if a court) that sources of this type are legitimate.

If seen as just a change in citation practice, this might be at best an interesting shift in the form of judicial opinions. But these changes do not merely reflect a transformation in citation practice. They embody a change in what counts as a legal source, and thus in what counts as a legal argument; and what counts as a legal argument-as opposed to a moral, religious, economic, or political one-is the principal component in determining just what law is. The boundaries of law are set by the boundaries of legal authority, and law speaks as law through its sources. When previously prohibited authorities become optional, and when previously optional authorities become mandatory, the nature of legal sources has changed, and with that change comes a transformation in the nature of law itself.

A mandatory precedent will sometimes, by virtue of its authoritative status, block an otherwise preferred current decision. … Law's use of precedent thus differs substantially from law's use of analogy, for in the latter a previous decision is selected in order to support an argument now, while in the former a previous decision imposes itself to preclude an otherwise preferred outcome.

Yet although sellers of beer and sellers of guns may in some respects be similar, in others they are not. Beer and guns are, after all, very different, and so are the typical circumstances in which they are sold. So when there are obvious similarities and obvious differences, the lawyer who has chosen one analogy rather than some other is relying on something that makes the similarities relevantly similar. … Just as the lawyers for the Nazis argued that a judge who thought the civil rights demonstration cases had been rightly decided ought to rule for the Nazis because of the relevant similarities between the cases, the lawyers for Skokie argued that even a judge who thought the civil rights cases had been properly decided had ample grounds to refuse to analogize those cases to one involving Nazis because of the presence of relevant differences.

In Adams, the question was the nature of the responsibility of the owner of a steamboat with sleeping compartments to an overnight passenger whose money had been stolen when, allegedly through the company's negligence, a burglar broke into the plaintiff's sleeping compartment. What makes the case interesting is that two divergent bodies of law were both potentially applicable. If the law pertaining to open sleeping compartments in railroad cars was applicable, then the company would not have been liable, but if the law pertaining to innkeepers applied, then the plaintiff would have been able to recover. The question was then whether the steamboat with closed passenger cabins was more like a railroad sleeping car or more like a hotel. … But in deciding that the steamboat was more like a hotel than a railroad sleeping car, the response to the skeptical challenge would go, the court did not first imagine what, on the basis of policy or principle, the best rule would be and then determine similarity on that basis. Rather, it looked at the two possibilities and simply "saw" more of a similarity in one direction than another. This might in theory have been reducible to some rule, but the rule did not consciously exist in the minds of the judges at the time they identified the similarity, so in fact for the judges the identification of similarity was a primary mental activity.

If the determination that a result is bad or in some way suboptimal is based on the full range of considerations of policy and principle that the court or anyone else would use without the rule, then what work is the rule itself doing? If our ability to perceive that a rule has generated the wrong result necessarily requires a sense of what the right result would be, and if that determination cannot be made by referring to the very rule whose soundness is now in question, then the common-law judge is deciding what the right result would be under a different rule or under no rule. And if every case in which the existing rule generates a result other than the one that appears correct in a rule-free evaluation is one that occasions a change in the rule, then why do we not just dispense with the rules entirely?

Whether Bentham was right, however, depends on whether the existing rule, true to the very idea of a rule, is entrenched firmly enough to resist or tolerate at least some suboptimal results. … If common-law rules will be modified when they produce very bad or substantially suboptimal results, but not when they produce only somewhat or slightly suboptimal results, then the rules will operate as rules do, even if they allow at times some number of somewhat or slightly bad or moderately suboptimal results.

Llewellyn's claim was less extreme than it may appear. As early as 1930, in The Bramble Bush, Llewellyn was at pains to limit his claims to the class of cases that were worth fighting over. Anticipating important law and economics insights about the so-called selection effect that were not to come until some fifty years later, Llewellyn recognized that the straightforward cases in which the law is all on one side rarely generate litigation. Indeed, they rarely generate legal disputes at all. Many Americans would prefer to pay their taxes on a date somewhat later than the April 15 deadline, but the implausibility of finding legal support for that position means that the question whether "April 15" in the Internal Revenue Code means April 15 will rarely be disputed, even more rarely be litigated, and more rarely yet wind up in an appellate court. Similarly, in the normal course of things, bills get paid, police officers obtain warrants, contracts are honored, and insurance companies whose insureds cause accidents make payments to the victims. Law abounds with such straightforward applications-we can call them "easy cases"-and the set of cases that winds up in court, and even more the smaller set that winds up in an appellate court, consists pretty much only of those cases in which both sides think that they have a colorable enough legal argument that it is worth spending time and money to go to court and then, for the losing party, worth the time and money of pursuing an appeal.

It would be a mistake to assume, however, that what is true for the Supreme Court is true for other courts and other issues. There has been somewhat less research focused on state courts and lower federal courts than there has been on the Supreme Court, but the body of that research is still considerable. And when we look at the conclusions of that research, we see that legal doctrine appears to play a considerably larger role in judicial decision-making than the more extreme of the Realists supposed. Although the self-reporting of judges probably exaggerates the effect of formal law on their decisions, the admittedly oversimplified conclusion that emerges from the research is that even in lower courts a range of nonlegal factors plays a larger role than the traditional model supposes, but that legal factors explain considerably more of lower court than of Supreme Court decision-making.

Although the empirical analysis of Supreme Court decision-making has become increasingly technical and sophisticated, the empirical claims of the Realists are essentially agnostic as to method. Some of the Realists urged research that employed what were at the time the cutting-edge methods of the social scientists, and would thus likely have been sympathetic to what are now the more sophisticated methodologies. But others believed that the careful perceptions of experienced lawyers would be sufficient to identify the "real" determinants of judicial outcomes and the real divisions or categories of the law.

It is worthwhile repeating, however, that the empirical assessment that the Realists have urged may in fact turn out for some courts and some issues and some types of law to be less inconsistent with the traditional view of law than most of the early Realists imagined. It may well be, for example, that the principal determinant of judicial decisions on questions of statutory interpretation is the ordinary meaning of the words of the relevant statute, and that the chief determinant on questions of contract law in appellate cases is the traditional rules and principles and doctrines of contract law as found in conventional contracts casebooks and in treatises like Corbin and Williston. Such outcomes would have surprised the Realists, but such traditional legal explanations for judicial outcomes may well be sound for some or many domains, and the very fact of taking this to be an empirical question is, in the largest sense, perhaps the most important feature and legacy of the Realist program.

A common saying is that "we are all Realists now," but that is almost certainly false. Not only is the "all" an egregious exaggeration, but the form of Realism that survives turns out to be a highly domesticated one. Beliefs in the total determinacy of legal doctrine may have withered, but torts casebooks look less like Leon Green's than might have been predicted half a century ago, with the typical modern book relying heavily on the traditional legal categories of tort doctrine. Constitutional law is still largely discussed, argued, and organized in substantial disregard of what the attitudinalists have rather firmly established, and any student who thinks that a strong Realist perspective will be rewarded on law school examinations is in for a nasty shock.

Judges created much of American antitrust law on a blank slate, for example, when they were forced to interpret the Sherman Antitrust Act of 1890, a statute whose main operative provision simply prohibits "[e]very contract, combination, ... or conspiracy, in restraint of trade or commerce."' In enacting such a law, Congress's use of imprecise language was not a matter of carelessness in drafting. Congress plainly knew what it was doing, and it knew how to use narrow and precise language when it wanted to. In writing the Sherman Act in broad and indeterminate language, therefore, and in thus intentionally avoiding concrete language and easily understood rules, Congress was instructing the courts to create, in commonlaw fashion, pretty much the entire body of antitrust law.

Fuller's misguided foray into the philosophy of language, ironically, detracted from rather than supported his highly valuable central point. The war memorial made out of a functioning military truck really was a vehicle, just as the tired businessman really was sleeping in the station, and just as the sheriff in Kirby really did obstruct the delivery of mail. What these and countless other examples, both real and hypothetical, show is that the application of the literal language of a rule will now and then produce an outcome that is absurd, ridiculous, or at least at odds with the principal purpose lying behind the rule.

The debates about the permissibility (or necessity) of recourse to legislative intent when a statute is unclear should not be confused with arguments about the purpose of a statute. It is legislators (or their equivalents) who have intentions, but statutes can have purposes, and it is often possible to determine the purpose of a statute from the words of the statute themselves. Sometimes, of course, the statute will say what its purpose is, a phenomenon described (and praised) by Karl Llewellyn as a singing reason, his term for a statute that not only has a purpose but that also announces it loud and clear.

Where this latter view prevails, where judges are trusted to pursue reason even if occasionally they get it wrong, it is best to understand the literal interpretation of a statute as defeasible, a term we encountered in exploring the common law's methods in Chapter 6. The term, which originally comes from property law and is now frequently found in jurisprudential writing, suggests that there are some circumstances in which a rule or principle or legally indicated outcome might be defeated. In the context of statutory interpretation, therefore, the view would be that the literal interpretation is still the standard and still the approach in the first instance. But not only when the literal interpretation is absurd, but also when the literal interpretation yields an outcome inconsistent with common sense, or inconsistent with probable legislative intention, or inconsistent with the statute's purpose, the judge may depart from literal meaning in order to produce the most reasonable result.

Once we understand judicial opinion-writing as the practice of offering legal support for decisions often involving elements other than the formal law-justice, for example, or efficiency, or wise policy, or even mercy or pity-we can see that many, maybe even most, judicial opinions are not fully candid. They read in the language of legal or rule-based motivation or compulsion, but their goal is to provide support or justification for decisions reached for different reasons. Perhaps it would be good if this were otherwise. Perhaps judicial opinions should describe accurately how the judge reached the decision she did. But as long as the idea of a judicial opinion as the accurate report of a decision-making process remains far more an aspiration than reality, the alleged importance of judicial candor turns out to be a substantial exaggeration.

Thus, when a court provides a reason for a decision, it gives a justification necessarily broader than that decision, and accordingly it announces what is in effect a rule (or a principle, standard, norm, or maxim) more general than the decision itself. To provide a reason in a particular case is to transcend the very particularity of that case. And indeed, the same structure operates when a court seeks to justify a rule or principle itself. Just as providing a reason for an outcome ordinarily takes the outcome to a greater level of generality, so too does providing a reason for a reason-or a reason for a rule or a reason for a principle.

It is an important consequence of the generality of reasons that a person (or a court) who gives a reason for a decision is typically committed to that reason on future occasions. If I tell a friend that I give money to Oxfam because it helps provide food to starving children in Africa, it is not at all surprising when my friend then asks me whether I will give money to his own pet charity, which also provides food to starving children in Africa. I may at that point be able to draw a distinction-the two organizations might be different in some other respect, or I might just be out of money-but having given the reason in the first "case," I am at least presumptively committed to following it in subsequent ones.

By way of contrast to open-ended legal terms like "best interests" and "unreasonable," consider the specificity of the regulation under the Occupational Safety and Health Act requiring that on all construction sites with "[m]ore than 20" and "fewer than 200" employees there shall be no less than "[o]ne toilet seat and one urinal per 40 employees." In much the same way, a Securities and Exchange Commission rule promulgated under the Securities Act of 1933 directs registrants to file "three copies of the complete registration statement" on "good quality, unglazed, white paper no larger than 8/ x 11 inches in size." And Article III of the Constitution mandates that "[n]o Person may be Convicted of treason unless on the Testimony of two Witnesses to the same Overt

It is true that location on the rules-standards continuum is an important way of allocating discretion between the issuer of the directive and those who must apply it, enforce it, or interpret it. But the difference between rules and standards is also a way of allocating decision-making between the present and the future. When a legislature, agency, or court sets forth a rule, it is making a decision now about what is to be done in the future. And when instead it moves toward the standards end of the continuum, it holds things open for the future and allows for a flexible approach to the problems of tomorrow.

A court concerned with not deciding too much but also concerned with giving guidance might, for example, make a decision containing a precise but narrow rule. The Supreme Court might still have specified, as in Miranda, close to the exact words that a warning would have to contain, but might have limited, as it did not, the cases to which those words would have to be given to a narrow class of police interrogations or to a narrow class of crimes. In such case, it would have selected an approach to rulemaking that was on the rules end of the rules-standards continuum but on the narrow end of the broad-narrow one. And in doing so, it would have done something quite different from what was done in Miranda, where the Court set forth a rule that was, like the class of insects, both broad and precise.

Because law is committed to the distinction between the trier of fact and the determiner of law, findings of fact are typically separated from conclusions of law when the same trial judge takes on both tasks. More importantly, findings of fact are typically, except in the most egregious of instances, treated as sacrosanct in the appellate process. It is only slight hyperbole to say that if a jury were to find that the moon was made of green cheese, an appellate court ruling on a legal question about the moon or about green cheese would be expected to take the jury's false conclusion as true.

This empirical conclusion may well be true, but it is not at all clear how the Supreme Court knew that it was true. Some might think the proposition self-evident, but once we realize that uninhibited, robust, and wide-open press criticism of public officials exists in countries with far more restrictive defamation doctrines (Australia, for example) than exist in the United States, it becomes less clear that the factual proposition that provided the linchpin for the Court's conclusion was as selfevident as the Supreme Court thought it was.

Even when a judge does not cite to nonlegal academic journals or newspapers or anything else, she is still, although less obviously, relying on sources of information that are importantly factual, that may very well be contested, and that wind up being part of the law in a somewhat under-the-table manner, even apart from the way in which such propositions may produce adverse consequences for one of the parties without that party having much or any opportunity to challenge those propositions by the normal adversarial processes, including but not limited to cross-examination.

Citation to materials outside of the traditional legal canon may be for some a source of alarm, but it may as well be a way in which the empirical propositions that are necessarily a part of all judicial lawmaking and much judicial law application can be subject to argument and challenge, rather than simply being clothed in the disguise of common knowledge or what judges believe, not always correctly, and not necessarily unrelated to their own backgrounds, to be the common wisdom of humanity.

Such a presumption might seem unfair, and indeed this kind of presumption seemed very unfair to the Supreme Court for a few years back in the 1970s. But as the Supreme Court came to realize upon further reflection, an irrebuttable presumption is little different from any other legal rule. Yes, the possessor of 28 grams of cocaine might not be a drug dealer, but the person who drives at 70 miles per hour in a 55-miles-perhour zone might not be driving unsafely, and the insider who buys and then sells shares in her own company in a less-than-six-months period may not be in possession of any inside information at all. But, as we saw in Chapter 2, rules do their work precisely by cutting off access to their background justifications, and few are surprised that exceeding the speed limit while driving very safely is still an invitation to a traffic ticket.

In accepting the inevitability and strategic or long-term desirability of some number of mistakes of mostly the right kind, the legal system's use of burdens of proof, presumptions, standards of review, and principles of deference, perhaps especially the last, resembles the system's use of rules and precedents, and resembles the legal system's at least partial commitment to formality. It may also resemble the legal system's willingness to make decisions on the basis of less than all of the best or available information. In all of these dimensions, the law, more than many other decision-making institutions, commits itself to accepting wrong or at least suboptimal answers, and it does so in the service of larger or longerterm institutional values, as well as service to the idea that the best way to get the largest number of correct decisions in the long term is often something other than attempting to make the best decision on every occasion. In operating in this way, law and legal reasoning may not be different in kind from other decision-making institutions, but they may differ in degree. At the heart of much of law's use of its characteristic reasoning devices is its acceptance of the fact that the best decision is not always the best legal decision.

Because straightforward or easy applications of legal rules are rarely litigated, the cases that come to a court are predominantly and disproportionately ones that are in some way hard. The litigated hard cases thus represent a biased sample of all legal events, a phenomenon typically referred to as the selection effect.

A court giving reasons for its decision in one case is announcing reasons that will apply in subsequent cases. Because reasons are always more general than the outcomes that they are reasons for, courts will often try to assess whether the reasons they give for a good result in the first case will have the effect of producing less-than-good results in subsequent cases. And courts will sometimes consider reaching the wrong outcome in the case before it in order to avoid laying down a rule that will produce poor outcomes in future cases.

A word is ambiguous when it is susceptible to two (or more) quite distinct meanings, as when we are unsure whether the word "bank" refers to the side of the river or the place where we keep our money or whether a "vessel" is something into which we put water or something that floats upon it, but this is rarely an issue in statutory interpretation. In interpreting statutes or other legal texts, the problem is usually that the words have no clear meaning rather than one or another clear meaning, and the correct word for this phenomenon is "vagueness" and not "ambiguity."

It is worth stressing that "open texture" is not the same as vagueness, but is rather the characteristic of any language, even the most precise language, to become vague in the face of unforeseen applications. Open texture is not vagueness but is rather the omnipresent possibility of vagueness.

What is most interesting is that Fuller recognizes that there might be several different ways to avoid an unjust outcome. One is to interpret the statute in contradiction of its plain terms. But there are others, including holding the law to have been violated but imposing a minimal sentence, holding the law to have been violated but refusing to enforce the law, and imposing a sentence while urging the executive to pardon the offenders.

Fortescue, Blackstone, and Starkie were all expressing the same principle, but the differences among ten to one, twenty to one, and ninety-nine to one reflect different views about the comparative harms of the two types of errors.