Stanford Law Review

February, 1987




Frederick Schauer[FNa]


Copyright 1987 by the Board of Trustees of the Leland Stanford Junior University; Frederick Schauer


       What does it mean for a past event to be precedent for a current decision? And how does something we do today establish a precedent for the future? Can decisions really be controlled by the past and responsible to the future, or are appeals to precedent just so much window dressing, masking what is in reality a decision made for today only? [FN1] And even if precedent can constrain decisionmakers, why should a procedure for decisionmaking impose such a constraint? Why should the best decision for now be distorted or thwarted by obeisance to a dead past, or by obligation to an uncertain and dimly perceived future? Equally important is the question of weight. When precedent matters, just how much should it matter? If we are to be shackled to the past and beholden to the future, just how tight are those bonds, and what should it take to loose them?


       An appeal to precedent is a form of argument, and a form of justification, [FN2] that is often as persuasive as it is pervasive. The bare skeleton of an appeal to precedent is easily stated: The previous treatment of occurrence X in manner Y constitutes, solely because of its historical pedigree, a reason for treating X in manner Y if and when X again occurs. [FN3]


        *572 Appeals to precedent do not reside exclusively in courts of law. Forms of argument that may be concentrated in the legal system are rarely isolated there, and the argument from precedent is a prime example of the nonexclusivity of what used to be called ‘legal reasoning.’ Think of the child who insists that he should not have to wear short pants to school because his older brother was allowed to wear long pants when he was seven. Or think of the bureaucrat who responds to the supplicant for special consideration by saying that ‘we've never done it that way before.’ In countless instances, out of law as well as in, the fact that something was done before provides, by itself, a reason for doing it that way again.


       Reliance on precedent is part of life in general. The law's lack of an exclusive claim on precedential reasoning suggests that we should dig beneath the use of precedent in this or even any legal system. I would thus like to look at precedent, at least initially, without the baggage of particular legal doctrines and without the encumbrances of a theory of law or judicial decision. By removing these potentially distorting constraints, we may see things that have previously been blocked from view. And having abstracted the analysis by looking at precedent outside of the law, we may then be able to glimpse a different picture of the use of precedent in law, and of what it means to ‘do law,’ to operate in a specifically legal form.




       An argument from precedent seems at first to look backward. The traditional perspective on precedent, both inside and outside of law, has therefore focused on the use of yesterday's precedents in today's decisions. [FN4] But in an equally if not more important way, an argument from precedent looks forward as well, asking us to view today's decision *573 as a precedent for tomorrow's decisionmakers. Today is not only yesterday's tomorrow; it is also tomorrow's yesterday. A system of precedent therefore involves the special responsibility accompanying the power to commit the future before we get there.


       Thinking about the effect of today's decision upon tomorrow encourages us to separate the precedential effect of a decision from the canonical language, or authoritative characterization, that may accompany that decision. Looking at precedent only as a backward-looking constraint may produce a distorted preoccupation with the canonical statements of previous decisionmakers. The precedents of the past, especially judicial precedents, come neatly packaged, with selected facts and authoritative language. Dealing with the use of past precedents thus requires dealing with the presence of the previous decisionmaker's words. [FN5] These words may themselves have authoritative force, what Ronald Dworkin calls the ‘enactment force of precedent,’ [FN6] and thus we often find it difficult to disentangle the effect of a past decision from the effect caused by its accompanying words. More pervasively, even a previous decisionmaker's noncanonical descriptions channel the way in which the present views those past decisions. So long as the words of the past tell us how to view the deeds of the past, it remains difficult to isolate how much of the effect of a past decision is attributable to what a past court has done rather than to what it has said.


       Nonjudicial precedents of the past also arrive at the present carrying with them their original characterizations. Suppose a child is allowed to stay up past her bedtime in order to watch a television show her parents thought particularly educational. Two weeks later, the child again wants to stay up late to watch a television show. The child will probably argue, ‘Two weeks ago you said I could stay up late to watch an educational show. Well, this show is educational too, so you have to let me stay up late tonight.’ That argument moves discussion away from the decision to allow the child to watch a particular show to what the parents said in making the decision. And then the argument resembles one over application of a specifically formulated rule such as a statute. If the parents had set out a rule, in advance, that bedtime was 8 p.m. except when an educational show was on later than 8 p.m., the discussion would differ little from that in which a precedent is accompanied*574 by simultaneous explanatory language. This is precisely because such language will be treated later like a specifically formulated rule. [FN7]


       The passage of time compounds the difficulty of disentangling a precedent from its specific linguistic account, because the process of characterizing a decision does not end with its first formulation. We necessarily and continuously reinterpret the past as we proceed into the future. People other than the initial decisionmakers use and talk about, and in the process recharacterize, the decisions of yesterday. The story of a decision changes as it passes from generation to generation, just as words whispered from child to child do in a game of ‘telephone.’ Past decisions thus come to the present encrusted with society's subsequent characterizations of and commentary on those decisions. [FN8]


       At the moment when we consider the wisdom of some currently contemplated decision, however, the characterization of that decision is comparatively open. [FN9] There is no authoritative characterization apart from what we choose to create. In making a decision, we must acknowledge its many possible subsequent characterizations, and thus the many directions in which it might be extended. Yet despite this seeming indeterminacy of the future precedential effect of today's decision, awareness of the future effect of today's decision pervades legal and nonlegal argument. Lawyers and others routinely deploy a battery of metaphors—the slippery slope, the parade of horribles, the floodgates, the foot in the door, and the entering wedge are but a few—to urge decisionmakers to consider the future effect of today's decisions. Undergirding each of these metaphors is the belief that even an uncharacterized precedent can influence the future.


       Thus, only the precedents of the past, and not forward-looking precedents, stand before us clothed with generations of characterizations and recharacterizations. When we look at today as a precedent for the future, we remove the distraction of the canonical effect of simultaneous*575 explanatory language, and we can better understand how being constrained by precedent often involves something different from being constrained by specifically formulated normative language.




       We can use the past for many purposes, but not every use of the past involves reliance on precedent. Initially, we must differentiate an argument from precedent from an argument from experience. The first time a child sees a red coil on top of a stove he may touch it, but when he sees the red coil again he knows its dangers and keeps his fingers safely clear. When a physician sees a certain array of symptoms that in the past have indicated typhoid, she will probably diagnose typhoid when those symptoms again appear. And many judges would describe the process of sentencing as based substantially on this type of experiential reasoning.


       In each of these cases, a present array of facts similar to some previous array leads a decisionmaker to draw on experience in reaching a conclusion. But despite the superficial similarity of these cases to the use of precedent, the reasoning process here is different. When reasoning from experience, the facts and conclusions of the past have no significance apart from what they teach us about the present. The probability that the present will be like the past both determines and exhausts the value of the previous experience. If we believe that the current case ought to be decided differently, no purely precedential residuum remains in the calculus. Moreover, if we now believe that the previous decision was incorrect, we will completely reject the value of the experience. But if we are truly arguing from precedent, then the fact that something was decided before gives it present value despite our current belief that the previous decision was erroneous.


       An argument from experience need not be based on the experience of the present decisionmaker. Unwilling or unable to do as much thinking, looking, or testing as a previous decisionmaker, the current decisionmaker may choose to conserve present decisional resources by relying on the prior decisionmaker's experience: ‘If Cardozo decided this way, who am I do disagree?’ Although people may describe this form of reasoning as ‘precedent,’ it is significantly different from a narrower, more central notion of precedent as a norm limiting the decisionmaker's flexibility. When the choice whether to rely on a prior decisionmaker is entirely in the hands of the present decisionmaker, the prior decision does not constrain the present decision, and the present decisionmaker violates no norm by disregarding it. I want to view precedent as a rule of precedent, and not as a nonrule-governed choice by a decisionmaker in an individual case to rely on the prior decisions of *576 others. [FN10]


       If precedent is seen as a rule directing a decisionmaker to take prior decisions into account, then it follows that a pure argument from precedent, unlike an argument from experience, depends only on the results of those decisions, and not on the validity of the reasons supporting those results. That two plus two today equals four is true, but it does not become more true because two plus two equalled four yesterday. When the strength of a current conclusion totally stands or falls on arguments for or against that conclusion, there is no appeal to precedent, even if the same conclusion has been reached in the past. If precedent matters, a prior decision now believed erroneous still affects the current decision simply because it is prior. Only if a rule makes relevant the result of a previous decision regardless of a decisionmaker's current belief about the correctness of that decision do we have the kind of argument from precedent routinely made in law and elsewhere.


       In referring to past and present decisionmakers, I presuppose that, as institutions, the past and present decisionmakers are identical or of equal status. If a pure argument from precedent is valid, the earlier decision has a status that must be respected by a later decisionmaker, and that status is imposed by the rule of precedent independent of other status differentials. Lower courts, for example, are expected to respect the decisions of higher courts. But the hierarchical ordering of decisionmakers implicates considerations different from those involved when a decisionmaker is constrained by its previous actions as opposed to the orders of its superiors in the hierarchy. [FN11]


       A naked argument from precedent thus urges that a decisionmaker give weight to a particular result regardless of whether that decisionmaker believes it to be correct and regardless of whether that decisionmaker believes it valuable in any way to rely on that previous result. Shorn of any embellishments, the reasons for respecting an argument from precedent may seem far from obvious. But before turning to that issue, we must first deal with the logically prior questions of whether precedent can matter, and if so, how.




A. Rules of Relevance


       Reasoning from precedent, whether looking back to the past or ahead to the future, presupposes an ability to identify the relevant precedent.*577 Why does a currently contemplated decision sometimes have a precedent and sometimes not? Such a distinction can exist only if there is some way of identifying a precedent—some way of determining whether a past event is sufficiently similar to the present facts to justify assimilation of the two events. And when we think about the precedential effect in the future of the action we take today, we presuppose that some future events will be descriptively assimilated to today's.


       No two events are exactly alike. For a decision to be precedent for another decision does not require that the facts of the earlier and the later cases be absolutely identical. Were that required, nothing would be a precedent for anything else. We must therefore leave the realm of absolute identity. Once we do so, however, it is clear that the relevance of an earlier precedent depends upon how we characterize the facts arising in the earlier case. [FN12] It is a commonplace that these characterizations are inevitably theory-laden. [FN13] In order to assess what is a precedent for what, we must engage in some determination of the relevant similarities between the two events. In turn, we must extract this determination from some other organizing standard specifying which similarities are important and which we can safely ignore.


       A parent's decision to let a daughter wear high-heeled shoes at the age of thirteen is scarcely precedent when a son then asks to be permitted to wear high-heeled shoes at that age. But a parent's decision to let that daughter stay up until ten o'clock will be relied upon justifiably by the son when he reaches the same age. A judgment finding tort liability based on the ownership of a black dog is precedent for a judgment regarding the owner of a brown dog, but not for a judgment regarding the owner of a black car. This is so only because a principle, or standard, makes dogness relevant in a way that blackness is not. Consider who among Alan Alda, Menachem Begin, and Dave Righetti is most similar to Sandy Koufax. One is the same age, another has the same religion, and a third is employed in the same occupation. The same point about the role of theory in assessing similarity undergirds Holmes' facetious description of the ‘Vermont justice of the peace before whom a suit was brought by one farmer against another for breaking a churn. The justice took time to consider, and then said that he had looked through the statutes and could find nothing about *578 churns, and gave judgment for the defendant.’ [FN14]


       It should be clear, therefore, that only the intervention of organizing theory, in the form of rules of relevance, allows us to distinguish the precedential from the irrelevant. Precedent depends upon such rules. And these rules themselves are contingent upon both time and culture. [FN15] Holmes used his churn story to show that legal similarity is determined by broad and theory-soaked descriptions like ‘property’ rather than by the nature of the objects involved. Would we draw that conclusion today? The existence of distinct principles for some classes of goods (consumer goods and securities, for example) shows that the rules of relevance in Holmes' time are not necessarily those of today. Consider the difference between 1865 Mississippi and 1985 Michigan in determining whether a decision about the competency of a white physician as a witness was precedent for a case involving a black physician as a witness. The race of the witness, now considered as irrelevant in determining expert competency as height or weight, at another time and place virtually disposed of the issue. Precedent depends on rules, and those rules themselves depend on context.


       With rule-dependency and context-dependency in mind, let us return to the forward-looking aspect of precedent. What is it to hope or fear that a decision will establish a precedent for some other decision in the future? Why do some people fear that allowing restrictions on Nazis because they are Nazis will establish a precedent for restrictions on socialists because they are socialists, even though the distinction between a Nazi and a socialist is obvious, accessible, and easily justified? [FN16] To worry about a precedential effect in the future, or to worry that a specific future event will be analogized to today's case, presupposes some rule of relevance. Rarely does someone at the time of the first decision think that almost the exact same facts will arise again. Rather, the thought is that this decision will establish a precedent for some different array of facts, one that contains some points of identity with the array of facts currently presented. Imagine a faculty meeting considering a request from a student for an excused absence from an examination in order to attend the funeral of his sister. Invariably someone will object that this case will establish a precedent allowing students to be excused from examinations to attend the funerals of grandparents, aunts, uncles, cousins, nieces, nephews, close friends, and pets. [FN17] Implicit in this objection is a rule of relevance that treats death as a relevant similarity, ‘caring’ as a relevant similarity, and any distinction between siblings and other meaningful relationships as irrelevant. In *579 some form or another, this type of rule of relevance inheres in any assertion of similarity.


       A rule of relevance may also be explained as a choice among alternative characterizations. To use the same example, the student's sister is simultaneously a woman, a sibling, a relative, a blood relative, and one with whom the student has a ‘meaningful relationship.’ How the relationship is characterized will determine whether later cases will be classed as similar. Thus, one who worries about establishing a precedent, or for that matter one who wants to establish a precedent, has in mind a characterization that the future will give to today's action, some category within which tomorrow's decisionmaker will place today's facts. The issue is one of assimilation, how we will group the facts and events of our world. The power of precedent depends upon some assimilation between the event at hand and some other event. This grouping may come from a simple characterization that includes both events, such as the phrase ‘censorship’ to include restrictions on the speech of both Nazis and socialists. Or the assimilation of the two events may come from a more complex rule of relevance, such as the rule that enables people to group government assistance to New York City, Lockheed, Chrysler, and farmers while distinguishing all of these cases from loans to students and veterans. In this case there is still some ‘heading’ under which some but not all instances are likely to be grouped. Regardless of simplicity or complexity, some grouping of the two events under one heading remains necessary for the operation of a system of precedent. The task of a theory of precedent is to explain, in a world in which a single event may fit into many different categories, how and why some assimilations are plausible and others are not.


B. Categories of Decision and Categories of the World


       Identifying the central place of a rule of relevance is only the first step. We must still locate the source of the rules enabling us to call something similar to something else. Only by looking to the source can we determine if that source constrains, or if a decisionmaker may classify the black dogs as easily with the black cars as with the dogs of other colors.


1. The articulated characterization.


       In seeking to locate the sources of characterization, we can take a large first step by noting the important distinction between decisions containing and those not containing canonical language. At times a decision will be accompanied by an articulated and authoritative characterization of the decision and its underlying facts. This articulated characterization, not unlike an articulated and specifically formulated rule, constrains the use of subsequent and inconsistent characterizations. Suppose the faculty grants the request of the student who wishes *580 to attend the funeral of his sister. Suppose further that the grant of the request is accompanied by a written explanation specifying that the excuse was given so that the student could attend the funeral of a member of his immediate family. This characterization of a sister as a member of an immediate family will in subsequent cases constrain (although by no means absolutely [FN18]) those who desire to characterize this precedent more broadly, such as by characterizing a sister as a relative, or as a close companion. A similar constraint would operate if the faculty that granted the first excuse attempted to narrow the precedent by denying an excuse in a subsequent case involving a father, mother, or brother. [FN19]


       In classical legal theory, articulated characterizations are often considered mere dicta, [FN20] but I do not want anything to turn on the questionable distinction between holding and dictum. Whether the characterizing language is treated as holding or dictum, that language cannot absolutely prevent a subsequent interpreter from recharacterizing the first case. [FN21] But that interpreter must at least confront an argumentative burden not present without an articulated characterization. [FN22]*581 Thus the articulated characterization acts like a specifically formulated rule. Rules can be broken, but to justify breaking a rule is more difficult than to take the same course of conduct in pursuit of or in the absence of a rule. Where there is an articulated characterization, therefore, the question whether precedent can constrain may collapse into the question whether rules can constrain. The issue is then the extent to which authoritative language can control, channel, or constrain decisionmakers purporting to interpret that language. This issue has been rehearsed for decades, [FN23] but it is different from the question whether a decision qua decision can also exert meaningful pressure on subsequent decisionmakers.


       The special problem of precedent thus appears most clearly when we remove the obscuring smokescreen of an articulated characterization. No decision in the real world is completely devoid of characterizations. Even a mere statement of facts is undoubtedly a characterization. Nevertheless, it is heuristically useful to imagine an event and a decision, without an authoritative statement of what the event is an example of. In this case, it appears possible for a subsequent decisionmaker to adopt any rule of relevance at all—or at least so many alternative rules of relevance that precedent appears to impose no constraint. Precedent's forward-looking aspect reveals this open-ended quality, for the forward-looking use of precedent is somewhat closer to the notion of a collection of facts without an authoritative characterization. Prior to a decision, no single authoritative characterization presents itself in the way that it does with a decision already made. Instead, one concerned about what the future might do with today's decision worries that this decision might be taken by the future to constitute a component of any number of categories extending this decision in different directions. When one fears that restricting Nazis may establish a precedent for content-based restrictions on other speakers, the fear is fueled precisely by the absence, at the time the fear is expressed, of any authoritative characterization separating Nazis from socialists. When faculty members fear the precedential effect of allowing some execuses from examinations, the fears are expressed in ways not present when specifically formulated rules are proposed, exactly because the absence of either rules or a canonical statement of the action in this case leaves open a wide choice of possible characterizations for the future to place on this decision. In any situation that lacks an articulated characterization, the central problem of precedent remains: Are there constraints on the assessment of similarity when few events are intrinsically similar to others? When there is no authoritative rule of relevance for a particular*582 situation, is it not then illusory to think of the precedent as in any way constraining?


2. The roots of characterization.


       We can now rephrase the problem. For any given decision, are there restraints on the possible categories of assimilation connecting the facts now before us with the facts before a future decisionmaker? Or is the choice of category completely up to the future decisionmaker, who can choose the category that a posteriori justifies the decision made on nonprecedential grounds? The problem is to determine what constrains a decisionmaker's control over the categories of assimilation.


       In exploring the question of a decisionmaker's control over the categories of assimilation, it may be useful to examine the philosophical controversy about categories, universals, and natural kinds. Under one view, the source of categories of assimilation is scarcely a problem at all. Philosophical theories that we can call ‘naturalist’ [FN24] argue that much of the world is divided into intrinsic categories, which the language of human beings merely recognizes. [FN25] Under this view, brown rocks differ from brown shoes because rocks and shoes are different. Period.


       The legal analogue of philosophical naturalism lies in one corner of that chamber of juristic horrors called ‘formalism.’ Under a rigid formalist [FN26] conception of precedent, rules of relevance track the natural and largely immutable patterns of the world around us, creating therefore no real choice [FN27] among rules of relevance on the part of either the *583 creator or the follower of a precedent. The only issue is the manner in which the naturally organized events are to be treated. The question of precedent in formalist theory is whether, having treated X in manner Y at one point in time, it was ever possible to treat X in manner Z at another. But to the formalist the only question is the comparative desirability of Y or Z. The question of why or how it is decided that both sets of facts are Xs never arises.


       Opposed to philosophical naturalism is the view commonly referred to as ‘nominalism.’ The nominalist views the world not as naturally subdivided, but rather as demarcated by the artificial labels contingently attached by particular people, particular cultures, or particular languages. Brown shoes and brown rocks are distinguished only because we draw the lines of our experience in certain ways rather than others. In a different world brownness rather than rockness might be more important, putting brown shoes in the same category as brown rocks but not in the same category as blue shoes. For the nominalist the world untouched by human beings contains only particulars, and universals, or general categories of any kind, are all impositions fashioned by people and their language.


       Nominalism, too, has its analogue in legal theory, and we call it Realism. [FN28] Taken with the wide range of characterizations available to a judge, many Realists saw the precedents of the past as invariably susceptible to whatever characterization best fit with the result the judge then wanted to reach. [FN29] Restrictions on Nazism and restrictions on socialism would be grouped together under the general rubric of ‘censorship’ only if the judge's theory of free speech condemned restrictions on both Nazis and socialists, or if the judge wanted both Nazis and socialists to speak. But if the judge liked socialists and not Nazis, then a different characterization—restrictions on racist speakers, for example—would enable the judge to drive a normative wedge between the two cases. These alternative characterizations were always available, according to the Realists, to enable a decisionmaker to mold the precedents of the past to the needs of the present. And if we in the present could so easily mold the past to conform to our current desires, then so too would the present be malleable in the hands of our successors. For some of the Realists, most notably Jerome Frank, the lesson *584 was clear: Judges should make their decisions for today alone. [FN30] What seems best in this case, considering all relevant factors with respect to these parties, is what should be done, without any sense of obligation to the past or responsibility to the future.


       Realism and formalism, like naturalism and nominalism, are caricatures. The world and the problems we confront in dealing with it lie between these polar extremes. But drawing the caricatures helps illustrate the problem—figuring out how categories of decision relate to categories of the world. And this analysis, in large part, will at least relate to, if not exclusively depend on, some broader outlook about where categories of different types come from, how they change, and how much they resist being manipulated by individuals.


       Naturalists and nominalists debate the question of whether the categories of the world come to us in relatively immutable form or whether they are largely the product of the minds and languages of particular cultures. The debate in the context of precedent is parallel but at a different level. The question is not whether there are immutable categories of the world. It is whether the categories used in making a decision are totally within the control of a decisionmaker, or of the decisional environment in which individual decisionmakers operate, or whether categories in some way resist molding by the decisionmakers who use them. When we see the issue in these less grand terms, the case for a naturalist/formalist perspective seems more plausible than it might otherwise seem. Even if nominalist/Realist perspectives do map the long-term workings of language and society, naturalist/formalist perspectives may nevertheless better describe the immediate actions of individual decisionmakers within a given social subculture.


       Our nonlegal world inevitably involves and incorporates its own categories of assimilation, its own rules of relevance. Think about why power tools are sold in hardware stores rather than in electrical appliance shops. And think about why we most often group red bicycles with bicycles of other colors rather than with red ties and red meat. Without such classifications our existence would be an undifferentiated sensory bombardment, akin to six radio stations broadcasting on the same frequency.


       Many of these larger rules of relevance are implicit in the structure of our language. Despite obvious differences, we refer to both amateur chess and professional football as ‘games.’ [FN31] Despite obvious similarities,*585 we have a different word for killing deer in the woods in Vermont than we have for killing cows in the stockyards in Omaha.


       Although there is a raging philosophical dispute about whether there are natural classes of the world, many of the nonultimate social categories with which we deal daily have clearly been created by human beings, vary with time and place, and are, like glaciers, constantly in flux despite their apparent immobility. People five feet six inches in height are ‘tall’ in some parts of the world and ‘short’ in others. Placing a shoplifter in the stocks is likely to be considered ‘cruel’ in 1986 in a way that it was not in 1791.


       The temporal and cultural contingency of the rules of life and language is not controversial. But it does not follow from the temporal and cultural contingency of rules of relevance that they are subject to change by a given decisionmaker, or even by a particular decisionmaking institution. That a society may change its collective views about what counts as a ‘vehicle,’ or what is ‘liberal,’ ‘fair notice,’ or ‘cruel,’ does not mean that the power to accomplish any of these changes exists in the short run, or inheres in individual or institutional decisionmakers. Take for example the question of the student asking to be excused from an examination in order to attend the funeral of a sibling. Would this establish a precedent with respect to cousins as well as siblings? The question cannot be avoided by noting that a future decisionmaker will have to make that decision. The question is whether there is some preexisting linguistic or social category that will group siblings and cousins together for some purposes, and, if so, what effect the existence of that category would have on the future decisionmaker's decision. So too with the possibility that some larger and preexisting category of assimilation will group restrictions on Nazis with restrictions on socialists despite the existence of a plausible distinction between the cases. If there is no such larger category of assimilation, there may be no risk that the desired decision of today will be grouped tomorrow with a decision that we today see as both undesirable and distinguishable. [FN32] But if such a category of assimilation exists in the larger consciousness surrounding the particular decisionmaking individual or institution, then there will be significant resistance to any decision that would disregard that category.


       As noted above, decisionmaking within the legal structure is constrained*586 by the comparative fixity of those larger societal and linguistic categories that help to constitute the conceptual apparatus of lawyers and judges. Legal decisionmakers, in purportedly relying on precedent, presuppose certain rules of relevance, and these rules of relevance are unquestionably contingent and subject to change. But if these rules of relevance come to legal decisionmakers from the larger linguistic and social environment, we would be mistaken to call them contingent within the legal decisionmaking subculture. [FN33] My point is that there is a profound difference between absolute contingency and contingency relative to the decisions then available within a particular domain. Legal rules of categorization are like passengers sitting on a train. In an important way the passengers are moving, but in an equally important way the passengers are sitting still.


       As an example of the dependency of legal decisions on the language and conceptualization of the larger world in which law operates, consider Federal Baseball Club v. National League. [FN34] Implicit in that case is the question of whether professional baseball will be assimilated with other members of the larger category ‘local exhibitions' or assimilated with other members of the larger category ‘businesses in interstate commerce.’ Even accepting the Court's dubious premise that one event cannot be a member of both classifications, the decision to put professional baseball in the ‘exhibitions' category deserves closer inspection. Nothing in the law compelled this result. It was instead influenced by the fact that the larger social and linguistic universe within which the Court operated assimilated a baseball game and a Fourth of July parade into the same category, and thus treated the business of selling entertainment and the business of selling oil as distinct. I am not saying the Court was correct, nor am I adopting the ‘formalist’ view that the law must always incorporate the definitions and distinctions then current in ordinary language. But I am saying that the law speaks largely if not exclusively in English, and relies significantly on social and linguistic categories drawn from its larger environment. [FN35]*587 However contingent these categories may be in that larger culture, a court at least bears an argumentative burden if it wishes to depart from these categories. And although I have used an example from law, the point applies to any decisionmaking environment operating within a larger social and linguistic culture.


       Thus, the central question involves the degree of control any particular subculture has over these larger categories. In the law, for example, how many of the theories that create theory-laden categories are legal theories, theories open to the legal culture to change on its own initiative? Some theories and categories of the law are indeed open to such change, and in those cases a theory of precedent within law relies upon a narrower explanation of how purely legal rules are made and changed. [FN36] But many other rules of relevance come from the larger world, and, to the extent they are relied upon in framing precedent, the law is inevitably constrained in its short-term ability to alter its own starting points. Precedent rests on similarity, and some determinations of similarity are incontestable within particular cultures or subcultures. If we can view these similarities as subculturally fixed, or subculturally incontestable, while at the same time recognizing their larger contingency, we can recognize that the rule-dependency and context-contingency of precedent do not force the conclusion that all characterizations of a past event are always up for grabs. If the available characterizations constitute a closed and often rather limited set, requiring a decisionmaker to follow some earlier result can substantially affect decisions.


       Thus, it seems simply mistaken to argue that the rule-dependency of precedent compels the conclusion that reliance on precedent collapses analytically into reliance on substantive normative rules. [FN37] The notion of precedent largely depends upon the way in what the categories of the world are frequently larger or more powerful in channeling thought than the categories or comparisons that might be crafted to particular normative purposes. The categories of our existence, the categories that group beanbag chairs and metal desks together as ‘furniture’ and motorcycles and trucks together as ‘vehicles,’ transcend specific normative ends. These conceptual categories intrude into diverse places and presumptively create assimilative groupings even where that assimilation may frustrate institutional or individual goals. The grouping of trucks and motorcycles as vehicles is a rule of language, not solely *588 within the control the regulations of the park department. We must distinguish the rule of language from the rule about the kinds of things we want in the park. If we collapse this distinction, we ignore what is most important—the way in which larger categories and rules of language that generate certain groupings significantly constrain the substantive rules in particular regulatory contexts. [FN38]




       I have argued to this point only that precedent can matter, that we may coherently talk about one decision as constraining precedent for another. Sometimes an articulated characterization or an articulated rule of relevance may constrain significantly the range of a category of assimilation. And even without an articulated characterization, the characterizations implicit in our organization of the world, although themselves mutable, still channel the determination of relevant similarities within a decisionmaking institution and narrow the range of possible later analogies. That a system of precedent can be constraining says nothing, however, about either the desirability or the degree of those constraints.


       In order to examine the justifications for a system of precedent, it is useful to begin by spending some time examining the consequences of adopting such a system. The most obvious consequence, of course, is that a decisionmaker constrained by precedent will sometimes feel compelled to make a decision contrary to the one she would have made had there been no precedent to be followed. Although this obvious consequence is no less important for its obviousness, and although I will discuss issues relating to this consequence in the following section, I want to concentrate now on the forward-looking aspect of precedent, and consider how this angle on precedent reveals consequences not quite so obvious.


       A rule of precedent tells a current decisionmaker to follow the decision in a previous similar case. Using the language I have been employing here, we can say that a current decisionmaker is told to follow the decision of a previous case involving assimilable, if somewhat different, facts. But of course the current decisionmaker of today is the previous decisionmaker of tomorrow. Although it may seem counterintuitive, this fact causes the current decisionmaker to be constrained by precedent even if there has been no prior decision.


        *589 Even without an existing precedent, the conscientious decisionmaker must recognize that future conscientious decisionmakers will treat her decision as precedent, a realization that will constrain the range of possible decisions about the case at hand. If the future must treat what we do now as presumptively binding, then our current decision must judge not only what is best for now, but also how the current decision will affect the decision of other and future assimilable cases. Thus, the current decisionmaker must also take into account what would be best for some different but assimilable events yet to occur. The decisionmaker must then decide on the basis of what is best for all of the cases falling within the appropriate category of assimilation. In making this calculation, the decisionmaker will discount the future for the possibility that the future decisionmaker will disregard or distinguish today's decision. This discount cannot be complete, however, unless there is some reason to believe that future decisionmakers will feel no pull of precedent. But where future decisionmakers can be expected to feel the pull of precedent, today's conscientious decisionmakers are obliged to decide not only today's case, but tomorrow's as well. If the best solution to today's case is identical to the best solution for tomorrow's different but assimilable facts, then there is no problem. But if what is best for today's situation might not be best for a different (but likely to be assimilated) situation, then the need to consider the future as well as the present will result in at least some immediately suboptimal decisions.


       The constraints imposed by an obligation to the future can be illustrated numerically. Imagine a series of five decisions, each involving a choice between some variety of A and some variety of B, where A and B represent categories subsuming assimilable but in some respects different particulars, A(1) through A(5) and B(1) through B(5). And suppose that for each decision in a case of A(n) versus B(n), it is possible to assess the relative social value of different decisions based on either A or B. We can then sketch a series of five opportunities for descision:
















































































      If each decision were independent of every other decision, the optimal decisions for each case would be A in the first case (55 is greater than 45), B in the second (70 is greater than 30), third (60 is greater than 40), and fourth cases (55 is greater than 45), and A in the fifth case (60 is greater than 40), for a net positive value of +100 (+10, +40, +20, +10, +20) over the five decisions. But suppose instead that the decisionmaker in Case 1 knows that her decision will be followed in some array of future cases involving different particulars but within the same categories. Under these conditions the decisionmaker, knowing *590 that her decision in Case 1 will be followed in Cases 2 through 5, must effectively decide not only Case 1, but the other four as well.


       Where precedent matters within a system, the possibility of case-by-case optimization is no longer available, and what is left is the choice of deciding either for A(1 through 5) or B(1 through 5), a choice of -40 or +40. The rational decisionmaker would choose B, +40. That choice of B, at +40, although rational in light of the available alternatives, is still suboptimal in terms of the best result for each of the five cases (+40 is less than +100), and is also different from the choice that would be made in the first case if the decisionmaker could think about that case in isolation, and not as precedent for the future. So the choice of B, plainly the proper choice if precedent is to be taken seriously, is inferior when compared to the best results theoretically obtainable over a range of cases if each case can be looked at in isolation, and is also inferior when compared to the best result that this decisionmaker would reach in the case at hand. [FN39]


       Accepting the constraints of precedent thus entails taking into account an array of instances broader than the one immediately before the decisionmaker. And this in turn means that although in no case can we make a decision that is better than optimal for that case taken in isolation, in some cases we will make decisions that are worse than optimal for that case taken in isolation. It thus becomes plain that adopting a strategy of reliance on precedent is inherently risk averse, in the sense of giving up the possibility of the optimal result in every case in exchange for diminishing the possibility of bad results in some number of cases.


       Thus, if the faculty in the case of the student wishing to attend the funeral of a sibling believes that future decisionmakers will assimilate siblings with all other relatives, and if it also believes that in some significant number of nonsibling relative cases it would be wrong to grant the excuse, then the faculty may plausibly deny the request in the present case even though it believes that it would be right to grant the excuse in this case if the case were taken in isolation. When we look at things this way, we see the price of precedent plainly presented. To the extent that a conscientious decisionmaker takes into account the fact that this decision will, within a regime of precedent, substantially control future decisions with respect to similar but assimilable situations, then the conscientious current decisionmaker must partially make those future decisions as well. When this happens, there will be at least some cases in which the best decision within a regime of precedent is nevertheless a suboptimal decision for the case at hand, with each such case *591 being a concrete example of the costs of a system in which precedent matters.




       It should be apparent that the extent to which precedent constrains will vary with the size of the assimilating category. In other words, if the conclusions of one case apply to a sweepingly broad set of analogies (and encourage decisionmakers to make such analogies), then the constraints of precedent are likely to be substantial. Not only will a broad set of subsequent decisionmakers feel the impact of the original decision, but the original decisionmaker will feel a greater obligation in making a decision with such broad application. The bigger the group of cases the original decisionmaker is effectively deciding, the more constraining will be the mandate to treat all of those cases alike. Conversely, if the categories of assimilation are comparatively small, the decisionmaker need consider only a few cases beyond the instant case, and the constraints of precedent will be comparatively inconsequential.


       Thus, the extent to which precedent constrains will vary not only with a decisionmaking institution's dependence on categories of assimilation external to its environment, but also on the size of those categories. To some extent the question of the size of the institution's categories of assimilation will be a function of the size of the particular categories existing in the larger world. But the strength of the constraints of precedent will also depend on just how much variation of particular decisionmaker is encouraged or allowed to acknowledge.


       With respect to many decisions, a theoretically justifiable distinction between two events may be drawn if the full richness of individual variation is allowed to be a relevant factor. Distinguishing, for purposes of eligibility for a tax decision, a failing steel industry from a prospering robotics industry is hardly unjustifiable. So too might we distinguish in the examination excuse example the good student from the marginal one. Of course, such distinctions are not necessarily wise or just. My point is only that distinctions like these cannot be considered arbitrary. The central question, then, is how many nonarbitrary variations between events will a given decisionmaker be allowed to pursue. If every nonarbitrary variation is open for consideration, then precedent poses only an illusory constraint. But if, instead, decisionmakers must look to that tuncation of real world richness that comes from thinking about the world in large and rough groupings, then the constraints of precedent will be substantial.


       Because the constraint of precedent might not be an all-or-nothing affair, we must consider the way in which the size of the categories of assimilation might largely determine the strength of precedent. This is admittedly an odd way of thinking about weight, because it is more common to think of the question of size and the question of strength as *592 lying along different axes. [FN40] But when we turn to precedent this distinction between size and strength collapses.


       It might occur to us to say that precedent provides one reason for a particular decision without providing a conclusive reason. We can certainly imagine a system in which the historical treatment of X as Y provides for that reason alone a conclusive, irrebuttable argument for treating X as Y now. But it seems also that a relevant precedent can matter even if it has less weight. For precedent to matter in this nonabsolute way, the historical treatment of X as Y could be said to provide a reason for treating X as Y now. There might be other reasons for treating X as Y. More importantly, there might also be reasons for treating X as not-Y, or as Z. To say that precedent provides a reason for deciding in a particular way is not to say that following precedent is what we should always do, all things considered. [FN41]


       In thinking about the possibility that precedent might have less than controlling force, it may be helpful to look briefly at lawyers' usage. Lawyers are fond of talking about ‘binding’ precedent, but what constitutes bindingness needs closer analysis. [FN42] Some decisions, even within a given jurisdictional ambit, must be followed while others, hypothetically, might carry with them some lesser force. In this sense, a *593 binding precedent is one that must either be followed or distinguished. [FN43] This type of bindingness is most commonly associated with the effect of superior court decisions on inferior courts, but it may also occur in the subsequent treatment of an earlier decision by the same court. [FN44] We can also imagine prior decisions that create precedent, but that may have their precedential value outweighed by other factors. Precedent might in this sense be taken to be only presumptive. This approach gives a precedent necessary weight without making that precedent dispositive. Therefore, we can distinguish the idea of precedent as a constraint or a reason from the idea of precedent as absolutely binding or constraining.


       Although the idea of a presumptive but not absolute reason is sound, that idea may not work with respect to precedent, which in an important way is a special case. In other areas, when we say that a factor or consideration or even a right is overridden, we can describe the overriding event in terms different from those we use to identify the circumstances that create, in the first instance, the necessity of a special justification. Think about equal protection, for example, and that using a classification based on race creates a burden of justification on the state to show that it has a ‘compelling interest’ in employing such a classification. [FN45] Here the obligation of meeting a higher than normal standard of justification is generated by the existence of a classification based on race. But once that obligation is created, it may be possible, as in Korematsu, for the state to provide such a special justification, such as the exigencies of war, based on reasons that themselves cannot be described in terms of a classification based on race. Think about the example of telling a lie in order to, say, assist the apprehension of an escaped murderer. [FN46] The existence of a particularly strong justification does not make the lie any less a lie. The presumptive strictures against lying are overridden by factors not necessarily related, except in this instance, to lying. The act of lying and the act of apprehending murderers, just like the act of classifying by race and the act of successfully waging war, are described in different terms, and at this level of discourse they cannot be reduced to a common currency. The ability to distinguish coverage and protection, scope and strength, and therefore the ability to think about reasons as presumptive, is dependent on this feature—that the possible *594 overriding events are measured along a different plane than are those events creating the need for special justification in the first instance.


       Precedent lacks this feature. The constraints of precedent attach in those instances in which some set of events now before us is deemed to be relevantly similar to some set of events of the past. The condition creating the presumption is similarity. But any overriding condition can always be expressed in terms of the absence of similarity. We never decide not to follow a precedent just because we do not feel like following it, but because something about the instant case is different from the precedent case. And if this is so, then the burden-creating conditions and the burden-satisfying conditions are measured along the same plane—similarity and difference. It is as if the special burden of justification present in the case of racial classifications were capable of being overridden by showing that in this case the racial classification was not that bad, or was not really a racial classification. These are permissible arguments, but they are not arguments for overriding. They are arguments showing that the reasons for requiring an especially strong justification simply do not apply here.


       In some sense, then, the question of weight is a red herring, for a precedent is always followed or distinguished. [FN47] We never face a situation where a precedent presumptively ought to be followed, but some special overriding condition in this case leads us not to follow it. Rather, we say that this case is simply different—that there is actually no relevant precedent to follow or disregard. [FN48] When the present situation presents factors that ought to control the disposition, we can and will translate those factors into a distinction between this case and the precedent case. Thus, the set of cases in which a precedent is applicable but overridden appears to be an empty one.


       We must still address the question of how specifically we will describe the precedents of the past and the events of the present. Although it will always be possible to distinguish a precedent, this becomes comparatively harder if we describe and use the precedents of the past in general terms. If subsequent decisionmakers describe and carry on the relevant precedent as a case involving a ‘car,’ then an attempt to distinguish an instant case because it involves a red car will be unpersuasive, precisely because the breadth of the description in the *595 first case substantially limits possible distinguishing factors in subsequent cases.


       Thus, the question of strength once again dissolves into the question of size. Because we can always characterize any overriding event as a difference, it makes no sense to draw a distinction between whether a precedent is applicable and whether a precedent controls. Both inapplicability and noncontrollingness express themselves in terms of distinctions between an instant case and some putative precedent case. At the level at which we create the practices of precedent, the strength of the norms of precedent will be reflected in the generality of the categories in which decisions are made. A precedential category of ‘vehicles' will have greater force than smaller categories of ‘cars,’ ‘trucks,’ and ‘motorcycles,’ which in turn will have greater power over the future than categories of ‘Buicks' or ‘Toyotas.’ No precedential rule can begin to capture this question of size. We cannot say that the appropriate categories of decision are just this big, or just this small. Instead, the rules of precedent are likely to resemble rules of language—a series of practices not substantially reducible to specifics. A system in which precedent operates as a comparatively strong constraint will be one in which decisionmakers ignore fine but justifiable differences in the pursuit of large similarities. At first glance, a precedent-governed system described this way, as one in which relevant differences are suppressed, seems no more justifiable than one characterized as treating prior decisions as important merely because they are prior. So why would any society want such a system?




       I have looked at how precedent can constrain, at the strength of its constraints, and at the costs of such constraints in imposing a strategy of suboptimization. What remains is what to some might be the biggest question: Why should a decisionmaking mechanism incorporate substantial precedential constraints within it? Why should a decision in Case 1 constrain a decisionmaker in Case 2 and itself be constrained by the knowledge of this secondary impact?


A. The Argument from Fairness


       Among the most common justifications for treating precedent as relevant is the argument from fairness, sometimes couched as an argument from justice. The argument is most commonly expressed in terms of the simple stricture, ‘Treat like case alike.’ [FN49] To fail to treat *596 similar cases similarly, it is argued, is arbitrary, and consequently unjust or unfair. We achieve fairness by decisionmaking rules designed to achieve consistency across a range of decisions. Where the consistency is among individuals at the same time, we express this decisional rule as ‘equality.’ Where the consistency among decisions takes place over time, we call our decisional rule ‘precedent.’ Equality and precedent are thus, respectively, the spatial and temporal branches of the same larger normative principle of consistency.


       The idea of fairness as consistency forms the bedrock of a great deal of thinking about morality. Whether expressed as Kantian universalizability, [FN50] as the decisions that people would make if cloaked in a Rawlsian veil of ignorance about their own circumstances, [FN51] or simply as The Golden Rule, the principle emerges that decisions that are not consistent are, for that reason, unfair, unjust, or simply wrong.


       How can we apply this broad principle of fairness more specifically as a potential justification for adopting a decision procedure in which precedent matters? Initially, the principle that like cases should be decided alike would seem to make an unassailable argument for precedent. But the difficulty of denying that like cases should be decided alike is precisely the problem. The statement is so broad as to be almost meaningless. The hard question is what we mean by ‘alike.’


       Recall the discussion of the potential size of categories of assimilation. From the perspective of the size of the categories involved, the question is not whether like cases should be decided alike, for at that level of abstraction that norm would engender unanimous agreement. Rather, the question is whether the categories of likeness should be large or small. If the categories of likeness, of assimilation, are so small as to enable a decisionmaker to take into account virtually every variation between separate events, then like cases are indeed being decided alike, yet the norm of precedent scarcely constrains. But if relatively large categories act to group many slightly different particular cases under general headings of likeness, then the stricture of deciding like cases alike makes reliance on precedent a substantial constraint.


       The issue is thus not the sterile question of treating like cases alike. It is instead the more difficult question of whether we should base our decisionmaking norm on relatively large categories of likeness, or by contrast leave a decisionmaker more or less at liberty to consider any possible way in which this particular array of facts might be unique. The purely formal constraint of treating like cases alike does not speak to this question. Yet the first of these alternatives describes a system of precedent; the second describes a system in which the constraint of precedent is for all practical purposes absent.


       Alone, therefore, the argument from fairness, the prescription to *597 treat like cases alike, does not help us choose between a decisional system with a strong precedential constraint and one with virtually no precedential constraint. If we are to find arguments directly addressing the question of precedent, we must look for substantive reasons to choose larger rather than smaller categories of decision.


B. The Argument from Predictability


       The most commonly offered of the substantive reasons for choosing strong over weak precedential constraint is the principle of predictability. [FN52] When a decisionmaker must decide this case in the same way as the last, parties will be better able to anticipate the future. [FN53] The ability to predict what a decisionmaker will do helps us plan our lives, have some degree of repose, and avoid the paralysis of foreseeing only the unknown.


       As a value, predictability is neither transcendent nor free from conflict with other values. Yet predictability plainly is, ceteris paribus, desirable. We attain predictability, however, only by diminishing our ability to adapt to a changing future. [FN54] In the language of precedent, following a precedent at a particular time may product a decision other than the decision deemed optimal on the facts of the particular instance. Where this divergence is absent, the effect of precedent is minimal. But if following precedent will produce a result different from that which would be produced without the constraint of the rule of precedent, then we have identified a case in which precedent matters. And thus we can rephrase the question: To what extent is a decisionmaking environment willing to tolerate suboptimal results in order that people may plan their lives according to decisions previously made?


       When we reformulate the question in this way, it becomes clear that the force of the argument from predictability, even if persuasive in the abstract, will vary with numerous factors whose weight cannot be generalized across all decisional environments. For example, how often will the use of large generalizations prevent making decisions on the basis of unique facts that would be dispositive in a particular case? To answer this question requires delving into a panoply of factors relating to the kinds of decisions that are to be made in a given decisionmaking environment. These factors may include the size of the relevant categories, the likelihood of significant factual variation, and many others that vary from environment to environment. I thus merely note, without*598 intending to resolve, that one important issue is the expected frequency of suboptimal results.


       The consequences of such a suboptimal decision provide a closely related concern, one touched on earlier. Once we realize that maintaining a serious regime of precedential constraint entails some number of suboptimal decisions, we see that we can express the price we pay for predictability in terms of some number of decisions in which the predicted result, and therefore the actual result if precedent is followed, is not in fact the best result. But the relationship of these costs to the possible benefits of predictability will vary across different kinds of decisions. When Justice Brandeis noted that ‘in most matters it is more important that the applicable rule of law be settled than that it be settled right,’ [FN55] he was reminding us of one side of this question. The other side, of course, is that sometimes [FN56] it is more important that things be settled correctly than that they be settled for the sake of settlement. To take an extreme example, making all capital punishment decisions under a strict precedential rule would satisfy desires for predictability but would also entail putting to death some people who would live if their individual cases were scrutinized carefully. And, at the other extreme, many decisions involving the formalities of contracts or real estate transactions are decisions in which sacrificing optimality for predictability would involve negative consequences that are far from catastrophic.


       Finally, it is worthwhile adding to the equation some variability in the value of predictability. Much of what we value about predictability is psychological. I feel better knowing that the letter carrier will come at the same time every day, that faculty meetings will not be scheduled on short notice, and that April brings the opening of the baseball season. Predictability thus often has value even when we cannot quantify it.


       Thus, the value of predictability is really a question of balancing expected gain against expected loss. We ask how important predictability is for those affected by the decisions, and we then ask whether that amount of predictability is worth the price of the frequency of suboptimal results multiplied by the costs of those suboptimal results. But there is no best answer to this calculation, for the answer will vary with the kinds of decisions that given decisionmakers are expected to make.


*599 C. The Argument from Strengthened Decisionmaking


1. Decisionmaking efficiency.


       When a precedent has no decisional significance as a precedent, the conscientious decisionmaker must look at each case in its own fullness. But when a rule external to the decisionmaker compels reliance on the decisions of others, it frees the decisionmaker from these responsibilities. Although a decisionmaker in a precedential system may in every case consider whether to disregard those constraints, [FN57] this does not deny that a decisionmaker may choose to follow precedent, nor that a decisionmaker who chooses to rely on precedent is in most cases operating within the norms of the system. Thus, a decisionmaker choosing to rely on precedent may justifiably ‘relax,’ in the sense of engaging in less scrutiny of the case, where that decisionmaker chooses to rely on a precedent. And where a rule of precedent urges a decisionmaker to relax in this sense, the net product will be a substantial reduction in decisionmaking effort.


       In this respect, efficiency may justify a rule of precedent. [FN58] This argument properly relies on the fact that a regime of precedent allows less reconsideration of questions already considered than a system containing no rule of precedent. If we wish to conserve the decisional resources of the decisionmakers, therefore, we have an independent argument for a system of precedent.


       There is something curious, however, about the argument from decisional efficiency. Suppose we have a string of three decisionmakers making three assimilable decisions. Under a regime of precedent, the second decisionmaker must defer to the first, but her deference is based on nothing about either the first or second decisionmakers, because it is also the case that the third decisionmaker must defer to the second. If we consider such a system efficient or otherwise desirable, we must either perceive some actual similarity between the decisionmakers or believe that there is some reason to impose a presumed similarity where none in fact exists. When there is actual similarity, a system of precedent enables a decisionmaker to rely on the similar decisionmaker's conclusion about a similar case. Having found the facts in the instant case similar to some decision of the past, the instant decisionmaker need go no further, because the system of precedent assumes that the same result would have been reached even if the instant decisionmaker started all over again.


*600 2. Strengthening the decisionmaking institution.


       But now let us abandon the assumption of similarity among decisionmakers. If we retain the assumption of assimilability of events, more than mere decisional efficiency must be at work. The system of precedent must operate to dampen the variability that would otherwise result from dissimilar decisionmakers. Why should we encourage this process? One possibility is that it might be thought important to create the aura of similarity among decisionmakers even where none may exist. Using a system of precedent to standardize decisions subordinates dissimilarity among decisionmakers, both in appearance and in practice.


       Even more substantially, this subordination of decisional and decisionmaker variance is likely in practice to increase the power of the decisionmaking institution. If internal consistency strengthens external credibility, then minimizing internal inconsistency by standardizing decisions within a decisionmaking environment may generally strengthen that decisionmaking environment as an institution. [FN59]


       The considerations surrounding the argument from decisionmaking efficiency rest upon a broad notion about the value of stability in decisionmaking distinct from anything about the actual decisions made. It is also apparent that any attempt to stabilize decisionmaking in an unstable world is likely to produce some suboptimal results. The argument from enhanced decisionmaking, then, suggests that the efficiency advantages may justify putting on blinders to the full richness of human experience. This is by no means an implausible argument, but its strength with depend upon many of the same factors discussed in the context of the argument from predictability. Likewise, whether this strength is sufficient to outweigh its costs will vary with those same factors. Once it becomes clear that no argument from invariable principle supports either the argument from predictability or the argument from enhanced decisionmaking, the evaluation of these alternatives turns into a weighing of costs and benefits that varies with different decisionmaking settings. Further elaboration must be relegated to discussions *601 of the goals to be served and the characteristics of decisionmaking in various decisionmaking settings.


D. Precedent and Stability


       Although the various arguments for incorporating a rule of precedent into a particular decisional environment each retains its own irreducible core of justification, in some sense these arguments coalesce. Arguments premised on the values of reliance, predictability, and decisional efficiency all share a focus on stability for stability's sake. The extent to which stability will be promoted is largely a function of the size of the groupings employed within a decisional domain. We must ask whether, at the extremes, large numbers of differences between instances will be suppressed, thus generating comparatively few large categories, or whether all variation will be taken into account, thus generating many smaller categories. Naturally there are gradations between these extremes. But by viewing the issue this way, we see that the various abstract and formal justifications for precedent all rest on conclusions about what ranges of events we wish to treat in like manner.


       At times these size-determining factors may be based on substantive value choices. We may broadly draw the category of ‘person’ in many contexts precisely because we believe it substantively evil to distinguish among people on the basis of race, gender, national origin, or even height. At other times the size of the categories will be controlled by the size of the categories at large in the world. Our language and its accompanying conceptual apparatus often inhibit attempts by decisionmakers to draw fine distinctions among the particulars that inhabit the larger categories of our existence. No amount of stress on particularization, for example, could completely prevent categories like ‘people,’ ‘dwellings,’ or ‘vehicles' from hindering a decisionmaker's efforts to see differences among people, dwellings, or vehicles. Finally, and perhaps most importantly, the size of the categories of assimilation may be a function of what we expect to accomplish in a decisionmaking setting. This is particularly germane here, for it links the principle of stability, hovering around all of the separate justifications for precedent, to the particular system actually generated.


       Some decisionmaking environments emphasize today—the richness and uniqueness of immediate experience. In those environments we seek the freedom to explore every possible argument or fact that might bear on making the best decision for this case, for it is precisely the thisness of the case that is most vital. At its extreme, such a system might, and arguably should, deny the relevance of precedent entirely. The virtues of stability would bow to the desire ‘to get it just right,’ and in such a framework a past decision would have little if any precedential force. More realistically, perhaps, such a system might still acknowledge*602 precedent, but in small units. For if we see precedents as small units, full of rarely duplicated particulars, we are likely to find few cases in which the current small unit is like some small unit of the past.


       By contrast, other decisional environments focus on yesterday and tomorrow, emphasizing the recurrent rather than the unique elements of the human condition. Here precedent has its greatest role to play, generating a format for decisionmaking that channels decisions toward consideration of a comparatively limited number of factors likely to be repeated over time. In the context of this essay, this would translate into the use of larger categories of assimilation, gathering many conceivably distinguishable particulars within the embrace of the larger categories.


       Without a universal answer to the question of whether stability is a good thing, we cannot decide whether decision according to precedent is a good thing. Stability may be unimpeachable in the abstract, but in reality stability comes only by giving up some of our flexibility to explore fully the deepest corners of the events now before us. Whether this price is worth paying will vary with the purposes to be served within a decisional domain, and we get no closer to knowing those purposes by understanding the relationship between stability and categorial size. Still, focusing on this relationship is valuable, because it enables us to see more clearly just how stability is achieved and just what kind of price we must pay to obtain it.




       We can locate most of the foregoing within a picture of law, but a picture that differs in some respects from the usual one. Much of contemporary legal theory presupposes that a successful theory of law must situate the structural phenomenon of law within the framework of the political state. A theory of law that does not explain the difference between the laws of Spain and Texas, on the one hand, and the rules of etiquette, baseball, and private clubs, on the other, is treated as incomplete. I do not want to deny that there are important differences between France and the American Contract Bridge League, and that some or all of these differences might be theoretically examined and explained. But it is equally useful to stress the similarities rather than the differences—to try to show the strong link between the rules of a political state and the rules of a private club. And if the consequent theory of ruleness [FN60]—explaining the place of rules within all of these diverse entities—fails to distinguish among them, the theory has not failed, for drawing that distinction is not the task of a theory of rules. Just as an analysis of bones may not provide the genetic explanation necessary to distinguish Homo sapiens from cats, so too might a theory of rules not be *603 aimed at distinguishing the rules of the political state from the rules of entities with different functions.


       Implicit in this agenda, therefore, is a rejection of the view that a theory of law must identify some form of thinking or decisionmaking unique to legal institutions. No one subscribes to this ‘uniqueness thesis' in exactly this form, but it has certainly occupied a prominent place in contemporary legal theory. [FN61] Rejecting the ‘uniqueness thesis,’ as I do, may produce the conclusion that legal reasoning is present outside the courts, and that nonlegal reasoning is present within the judicial system. This conclusion may have uncomfortably fuzzy edges, but it makes up in accuracy what it lacks in theoretical elegance.


       Rejecting a strong version of the uniqueness thesis, however, does not mean that different forms of reasoning cannot be distinguished. That we cannot definitively situate the nature of law within the workings of the political state does not mean that some forms of decisionmaking do not differ from others. By distinguishing thinking that focuses on the present from thinking that mandatorily encompasses a wider temporal scope, I have tried here to begin exploring one way in which some forms of thinking may differ from others, and one way—the use of precedent—in which this difference may find expression in institutional design.


       Even the mention of institutional design may beg the important question, because it assumes, prematurely, that we should design institutions to fit one or another decisional model. It may be better to think in terms of decisional domains, recognizing that certain institutions may contain several such decisional domains working in parallel. For example, courts may at times rely on something close to the theoretical ideal of a system of equity, [FN62] while at other times they rely more on a rule- or precedent-based approach. When Congress considers a private bill, precedent is rarely mentioned, but when Congress impeaches and tries a president or a judge, [FN63] precedent rightfully comes into play. Similarly, the Department of Justice may rely less on precedent when establishing prosecutorial priorities than when giving advice to the president with respect to constitutional responsibilities.


        *604 Understanding the notion of precedent becomes easier when we focus on examples like these, because they force us to look at those settings where the constraints of precedent are largely absent. It would seem odd indeed to criticize Ronald Reagan for not following the precedent of Jimmy Carter with respect to relations with the Soviet Union. This is not to say that the policies of either man cannot be criticized. But the criticism must be substantive, and it is no independent ground for condemnation that a president has failed to follow the policy lead of his predecessor. Many aspects of the administrative process share this characteristic, and, like equity, at least one version of the ideal of the administrative process holds that an agency should be free to consider the fullness of certain problems without the constraints of following its or another agency's prior handling of a different array of highly complex facts. [FN64]


       Equity and administrative law are good examples of the fact that the ideals of rule-free and precedent-free decisionmaking, focusing on the richness of this case without taking on the burdens of the last case or the next case, are just that—ideals. But these ideals remind us that there is a difference between precedential constraint and total freedom to decide this case in the optimal way, even if the two processes rarely appear in pure form. Yet, by seeing that there is a difference, we see that the extent of reliance on precedent is not fixed, but is subject to change in the same ways as are norms governing any decisionmaking institution. This process does not necessarily depend upon written rules, but reflects the entire array of methods by which decisionmakers assimilate what it is proper to do and what is discouraged.


       Thus it may be that decisions for the moment, relatively unconstrained by precedent, are not alien to the courts, and that decisions attendant to precedent are not alien either to the political world or to the private worlds in which we live our lives. This in turn is part of a larger perspective that sees law, in this decisionmaking-process sense, in our nonlegal institutions, and that sees nonlaw, in this same decisionmaking-process sense, in our legal institutions. When we see this we may be entirely satisfied. But we may instead want to challenge the extent of this conflation of function. Perhaps we should view legal institutions, including lawyers and law schools, as part of a larger mechanism—call it society—that needs some institutions that are creative, speculative, adaptive, and risk-taking, and other institutions that are cautious, predictable, and risk averse. These latter institutions might act as stabilizers and brakes, rather than as engines and accelerators, and it may be that both forms of institution together constitute, or at least approach, the ideal mix of decisionmaking structures. Within this mix of structures it should be apparent that precedent, as an inherently *605 constraining form of argument, is more suited to some forms of decision than to others. And it should then be apparent why the constraints of precedent have been and perhaps should be reserved not for our institutions of progress, but for our institutions of restraint.


[FNa] Professor of Law, University of Michigan.

        Earlier versions of this article have been presented at faculty workshops at the University of Chicago, the University of Michigan, the University of Texas, the University of Toronto, the University of Virginia, and Yale University. The conventional form of thanks for these opportunities and attendant audience insights is insufficient to express fully my gratitude to the institution of the faculty workshop, which combines ideally the means for exploring one's ideas with others and the informality necessary for truly cooperative intellectual interchange. I would also like to thank Alex Aleinikoff, Larry Alexander, Edward Cooper, Patrick Gudridge, James Krier, William Miller, William Pierce, Thomas Rowe, Terrance Sandalow, Kim Scheppele, Carl Schneider, Peter Westen, and James Boyd White, whose willingness to take time from their own work to comment on mine has made this article better than it otherwise would have been.


[FN1]. Recall Justice Roberts' fear that Supreme Court decisions might come to be treated like ‘a restricted railroad ticket, good for this day and train only.’ Smith v. Allwright, 321 U.S. 649, 669 (1943) (Roberts, J., dissenting).


[FN2]. There is no difference between the logical structure of an argument and that of a justification, but the different terms suggest separate events within a larger rhetorical setting. Arguments are made prior to a decision, and thus are made to a decisionmaker. Justifications, on the other hand, are provided after a decision is made or an action is taken, and are given by the decisionmaker.


[FN3]. Standard treatments are found in R.W.M. DIAS, JURISPRUDENCE 162-217 (4th ed. 1976): G. W. PATON, A TEXT-BOOK OF JURISPRUDENCE 179-95 (D. Derham 3d ed. 1964); P. FITZGERALD, SALMOND ON JURISPRUDENCE 141-87 (12th ed. 1966). Traditional approaches with a more modern cast include C. K. ALLEN, LAW IN THE MAKING 161-382 (7th ed. 1964); R. CROSS, PRECEDENT IN ENGLISH LAW (3d ed. 1977).


[FN4]. A noteworthy exception is MacCormick, Formal Justice and the Form of Legal Arguments, 6 ETUDES DE LOGIQUE JURIDIQUE 103 (1976). Considering the future implications of today's decision is central to the notion of adjudication according to principle. See Golding, Principled Decision-Making and the Supreme Court, 63 COLUM. L. REV. 35 (1963); Greenawalt, The Enduring Significance of Neutral Principles, 78 COLUM. L. REV. 982 (1978); Wechsler, Toward Neutral Principles of Constitutional Law, 73 HARV. L. REV. 1 (1959). But principles, in and out of law, are expressed in words, and the existing discussion of principles involves the problem of imagining what future cases might later come within the domain now described by a particular set of words. That, as I will discuss presently, is distinct from the question of what later decisions might be taken to follow from the first decision. Implicit in what I say, therefore, is a distinction between what is said and what is done. Much of the discussion of principled decision-making focuses on being constrained tomorrow by what we say today. That is important, but it is not the same as being constrained by what we do today. Moreover, the focus of the ‘neutral principles' literature is on the commitments necessarily, or at least ideally, made by a single decisionmaker. That differs from the effect of a particular decision on some other decisionmaker possibly not sharing the same views.

More significantly, a careful study of precedent must confront the extent to which sticky and substantially nonnormative social or linguistic characterizations may impede the ability of a formulator of a principle to draw certain intrinsically sound distinctions or to employ certain intrinsically justifiable groupings. This problem, separate from the problem of decision according to principle, is central to understanding the concept of precedent.

[FN5]. This relationship among facts, facts as found by the court, decision of the court, and announced reasoning of the court is at the heart of the English debate about ratio decidendi. See Goodhart, Determining the Ratio Decidendi of a Case, 40 YALE L.J. 161 (1930); Goodhart, The Ratio Decidendi of a Case, 22 MOD. L. REV. 117 (1959); Montrose, Ratio Decidendi and the House of Lords, 20 MOD. L. REV. 124 (1957); Montrose, The Ratio Decidendi of a Case, 20 MOD. L. REV. 587 (1957); Simpson, The Ratio Decidendi of a Case, 20 MOD. L. REV. 413 (1957); Simpson, The Ratio Decidendi of a Case, 21 MOD. L. REV. 155 (1958); Simpson, The Ratio Decidendi of a Case, 22 MOD. L. REV. 453 (1959); Stone, The Ratio of the Ratio Decidendi, 22 MOD. L. REV. 597 (1959). In somewhat different language, it is also the centerpiece of E. LEVI, AN INTRODUCTION TO LEGAL REASONING (1948) and K. LLEWELLYN, THE BRAMBLE BUSH 51-66 (1930).




[FN7]. See Schauer, Opinions as Rules (Book Review), 53 U. CHI. L. REV. 682 (1966). One of the more important discussions of legal reasoning takes decision according to rule and decision according to precedent as identical forms of reasoning, and groups both under the rubric of ‘precedent.’ R. WASSERSTROM, supra note 6, at 56. To neglect the difference between decision according to a specifically formulated rule and decision according to an unformulated decision seems a distorting reduction. But even if one is inclined towards reductionism, taking decision according to precedent as a species of decision according to rule, see Western, On ‘Confusing Ideas': Reply, 91 YALE L.J. 1153, 1163-64 (1982), is a more plausible reduction than the reverse. Wasserstrom's terminology does, however, have the advantage of stressing the essential ‘pastness' of both rules and precedents. On this idea of pastness, see Lehman, Rules in Law, 72 GEO. L.J. 1571, 1573 (1984).


[FN8]. Think about the way in which cases ‘mean’ something different today than they did when first decided, especially when a case has passed from the legal into the larger culture. Brown v. Board of Education, 347 U.S. 483 (1954), ‘stands for’ something today to the public quite different from what it did to the public in 1954. It would be rather odd if this shift in meaning did not affect the understanding of this case in the legal culture as well.


[FN9]. I will presently argue that the available characterizations constitute a more limited set, but a case yet to be decided still presents a menu of possible characterizations larger than that of a precedent of the past.


[FN10]. A decisionmaker might choose to adopt a ‘rule of thumb’ to assist the process of deciding whether to rely on the experiences of others. But as long as the decision whether to adopt and follow the rule of thumb remains solely one for the decisionmaker, then it is not behavior that is rule-governed in the sense that it is guided by some external norm of behavior. See Urmson, The Interpretation of the Moral Philosophy of J. S. Mill, 3 PHIL. Q. 33, 34 (1953).


[FN11]. Note that while contemporary usage has collapsed much of the difference between precedent and stare decisis, it should be clear that I am talking about the latter in its strict sense.


[FN12]. A characterization may be a simple word or phrase, as when we characterize a 1957 Chevrolet as a car, a vehicle, or an antique. But a more complex description of what this event is an example of is still a characterization.


[FN13]. See, e.g., Shiner, Precedent, Discretion and Fairness, in LAW, MORALITY AND RIGHTS 93, 97 (M. A. Stewart ed. 1983) (if following precedent is to be logically derived, then the instant case and the precedent case must have ascertainable, legally relevant properties). Indeed, even less skeptical theorists about precedent agree on this. See, e.g., R. CROSS, supra note 3, at 182-92 (although in later stages judges are often misled in their reliance on precedent, the first stage in judicial reasoning by analogy is the determination of the relevant likeness between the previous case and the one before the court).


[FN14]. Holmes, The Path of the Law, 10 HARV. L. REV. 457, 474-75 (1897).


[FN15]. See Gordon, Critical Legal Histories, 36 STAN. L. REV. 57, 125 (1984).


[FN16]. I am, of course, referring to the discourse surrounding the American Nazi Party's effort to march in Skokie, Illinois: See generally A. NEIER, DEFENDING MY ENEMY (1979).


[FN17]. See generally Schauer, Slippery Slopes, 99 HARV. L. REV. 361 (1985).


[FN18]. I find it useful to think in terms of argumentative burdens, rather than in terms of absolute prohibitions. Imagine ordering something in a restaurant that is not on the menu. Ordering off the menu is by no means impossible, but it is certainly more difficult than ordering from the menu. Throughout this article I assume this presumptive rather than absolute sense of constraint. For a more substantial discussion, see Schauer, Easy Cases, 58 S. CAL. L. REV. 399, 423-26 (1985); see also Powers, Structural Aspects of the Impact of Law on Moral Duty Within Utilitarianism and Social Contract Theory, 26 UCLA L. REV. 1263 (1979) (discussing ‘translucency’); Powers, Formalism and Nonformalism in Choice of Law Methodology, 52 WASH. L. REV. 27 (1976) (discussing ‘translucency’).


[FN19]. The two situations are not exactly equivalent, and a sort of ‘ratchet effect’ seems to characterize the way in which constraints operate more in one direction than in another. Expansion seems more defensible than contraction, and it is important to think about why expansion of a given decision seems to flow more easily than contraction. A starting point may be the observation that we do not lead our lives under the expressio unius maxim. Except where something special about the context demands it, it is not a rule of discourse that what we say is all we have to say. But it is a rule of discourse that one does not willy-nilly go back on what one has already said. In some contexts, however, including most legislation and perhaps including case law, it is our understanding that what is said simultaneously says something and marks the outer limits of a domain. Thus the expressio unius maxim presupposes the existence of two groups—the group actually specified and the larger group, which could have been specified but was not. The question is one of deciding, given the absence of specification, what this larger group is. And this, in turn, will be largely a function of whether that larger group could have been specified. If there is no reason to suppose that it could not have been, then we are likely to take extension to that larger group in a later case as being as inconsistent with the first case as would be outright contradiction.


[FN20]. See R. CROSS, supra note 3, at 79-86. Compare Goodhart, Determining the Ratio Decidendi of a Case, 40 YALE L.J. 161 (1930) (the principle of law that is the stated basis of a decision is not binding precedent) with Montrose, Ratio Decidendi and the House of Lords, 20 MOD. L. REV. 124 (1957) (reason given for the decision is binding) and Simpson, The Ratio Decidendi of a Case, 21 MOD. L. REV. 155 (1958).


[FN21]. See E. LEVI, supra note 5, at 9-27. The locus classicus is K. LLEWELLYN, THE COMMON LAW TRADITION 85-91 (1960).


[FN22]. Even if there is a distinction between holding and dictum, the existence of dicta creates an argumentative burden that would not have been present had the first court said nothing at all. We commonly talk of ‘mere’ dicta as a way to justify ignoring some language in a case, but it is worthwhile to consider the effect of having to go through that exercise compared with the situation in which no such troublesome language exists.


[FN23]. See generally J. PAUL, THE LEGAL REALISM OF JEROME N. FRANK (1959); W. RUMBLE, AMERICAN LEGAL REALISM (1968); W. TWINING, KARL LLEWELLYN AND THE REALIST MOVEMENT (1973); Schlegel, American Legal Realism and Empirical Social Science: The Singular Case of Underhill Moore, 29 BUFFALO L. REV. 195 (1980).


[FN24]. Actually the standard label in philosophy for this perspective is ‘realism.’ This standard label, however, is best avoided here. The term ‘realism’ is distracting enough in a context in which Legal Realism is being discussed, but that term is doubly confusing in light of the fact that, as shall be shown, philosophical realism is the polar opposite of Legal Realism. This ought to give pause to those who too easily use terms like ‘formalism’ and ‘realism’ in place of case analysis. But it won't.


[FN25]. I ought to feel guilty about presenting in one paragraph the entire problem of universals, and doing so in a way that seems to make most of classical and modern metaphysics preposterous. But this is not an essay on metaphysics, and the relationship between categories of decision and categories of the world can best be presented by use of these admittedly crude descriptions of much more highly nuanced positions. Useful introductions to the philosophical views caricatured here include R. AARON, THE THEORY OF UNIVERSALS (2d ed. 1967); H. STANILAND, UNIVERSALS (1972); UNIVERSALS AND PARTICULARS (M. LOUX ed. 1970). For a sampling of contemporary discussion of these issues, see STUDIES IN METAPHYSICS (P. French, T. Uehling, Jr. & H. Wettstein eds. 1979).


[FN26]. It is unlikely that any person actually held in pure form the caricatured position that today is described as ‘formalism,’ although if one is searching for isolated passages the writings of Cesare Beccaria, William Blackstone, and Jeremy Bentham are fruitful terrain. Somewhat more extreme statements are found in E. WAMBAUGH, THE STUDY OF CASES (1894) and Zane, German Legal Philosophy, 16 MICH. L. REV. 287 (1918). I confess that I am using ‘formalist’ as a useful construct rather than as a characterization of the views of particular people.


[FN27]. In a relatively uninteresting sense, any actor always has a choice of whether to follow even a clear rule. In a more interesting sense, some rules, or some potential applications of rules, present an unavoidable choice. The question whether my bicycle is excluded by the ‘No vehicles in the park’ rule, see H.L.A. HART, THE CONCEPT OF LAW 123 (1961); Hart, Positivism and the Separation of Law and Morals, 71 HARV. L. REV. 593, 607 (1958), involves a choice of a different variety than the question whether, faced with the same rule. I my drive my truck in anyway. My reference in the text to a ‘real’ choice is a reference to a case in which a decisionmaker is faced with a decision among more than one plausible but mutually exclusive rule of relevance, and not to a case in which the decisionmaker as to decide whether to disregard the only plausible rule of relevance.


[FN28]. My disclaimer, see note 26 supra, about real formalists applies equally to the views of real Realists. Again, the caricature may serve as a useful heuristic device.


[FN29]. See, e.g., J. FRANK, LAW AND THE MODERN MIND (1930); K. LLEWELLYN, supra note 5, at 64-70; Douglas, Stare Decisis, 49 COLUM. L. REV. 735 (1949); Radin, Case Law and Stare Decisis: Concerning Präjudizienrecht in Amerika, 33 COLUM. L. REV. 199 (1933).


[FN30]. ‘Present problems should be worked out with reference to present events. We cannot rule the future.’ J. FRANK, supra note 29, at 155.


[FN31]. See L. WITTGENSTEIN, PHILOSOPHICAL INVESTIGATIONS 67-71 (G. Anscombe trans. 2d ed. 1958). As with the case of the word ‘games,’ I do not take the notion of a rule of language or of life as necessarily containing a specifiable list of necessary and sufficient conditions for the application of the rule. But I do take the categorizations of language and life as rules, in the sense that talking or acting in ways inconsistent with the categories (‘Bridge, elephants, and An American Tragedy are four of my favorite games.’) engenders a range of negative reactions that can be summarized under the heading of ‘breaking a rule.’ My view that some sort of rule is necessary for the operation of a system of precedent is thus consistent with Moore, A Natural Law Theory of Precedent, in PRECEDENT IN LAW (L. Goldstein ed. 1986), although we differ in a fundamental way on what we consider the source of rules significantly involved in the operation of a system of precedent. Contrasting ‘particularist’ views, effectively criticized by Moore, include S. BURTON, AN INTRODUCTION TO LAW AND LEGAL REASONING (1985); Christie, Objectivity in the Law, 78 YALE L.J. 1311 (1969); Stone, Ratiocination not Rationalization, 74 MIND 161 (1965).


[FN32]. I say ‘may’ because there are phenomena other than a shared assimilating category that may cause a decision made today to increase the probability of occurrence of some feared but different decision tomorrow. See Schauer, supra note 17.


[FN33]. It is also often the case, especially in law-laden countries like the United States, that the reverse also occurs, that the categories and concepts of the law influence thinking in the larger world around the law. See A.V. LUNDSTEDT, LEGAL THINKING REVISED 159-70 (1956); Freeman, Legitimizing Racial Discrimination through Anti-Discrimination Law: A Critical Review of Supreme Court Doctrine, 62 MINN. L. REV. 1049 (1978).


[FN34]. 259 U.S. 200 (1922); see also Flood v. Kuhn, 407 U.S. 258 (1972); Toolson v. New York Yankees, Inc., 346 U.S. 356 (1953).


[FN35]. The relationship between ordinary English and the English spoken, written, or interpreted in the legal culture is one of the most interesting potential fields of inquiry in legal theory, but one that has received remarkably little attention in the literature. On the one hand, it would be difficult to deny the extent to which legal language, as a technical language, often operates in a context that makes legal terms have meanings different from those they bear when spoken in other contexts. See Schauer, Speech and ‘Speech’—Obscenity and ‘Obscenity’: An Exercise in the Interpretation of Constitutional Language, 67 GEO. L.J. 899 (1979). On the other hand, in an important and largely unexplored sense, legal English must remain attached to ordinary English. It seems intuitively correct to note the way in which technical languages are parasitic or ordinary language. See Caton, Introduction to PHILOSOPHY AND ORDINARY LANGUAGE at vii-viii (C. Caton ed. 1963). But working this out for law, and taking into account the way in which law must simultaneously function for professionals within the legal system and for citizens outside it, is a task that remains to be completed.


[FN36]. See E. LEVI, supra note 5. Where Levi goes wrong, e.g., id. at 4, is in treating the categories used in law as necessarily legal categories, ones in control of the legal system.


[FN37]. Here I am responding to arguments made in, e.g., N. MACCORMICK, LEGAL REASONING AND LEGAL THEORY 163, 186 (1978); Westen, supra note 7, at 1163-64. A similar argument couched in terms of desirability rather than inevitability is Moore, A Natural Law Theory of Interpretation, 58 S. CAL. L. REV. 277, 358-76 (1985).


[FN38]. It is important to point out that the persuasiveness of my argument is likely to vary with the size of the domain we are contemplating. If we focus only on appellate cases, in which the system presents us exclusively with situations existing at the margins of rule and precedent, we will see many cases in which decisional constraints are inoperative and few in which they are effective. But if we recognize as well that precedent prevents many situations from being contested and makes other situations more difficult to contest, we will appreciate better the constraints that bind us even while they do not appear regularly on the litigator's agenda. See Schauer, supra note 18.


[FN39]. The model I use here is of course an extreme simplification, in part because it ignores the effects of precedents other than this decision, and in part because it assumes a monolithic decisionmaker. With regard to the latter assumption, see Easterbrook, Ways of Criticizing the Court, 95 HARV. L. REV. 802 (1982). I do not believe that either of the simplifying assumptions distorts my basic point.


[FN40]. See Schauer, Can Rights Be Abused?, 31 PHIL. Q. 225, 228-30 (1981) [hereinafter Schauer, Rights]; Schauer, Categories and the First Amendment: A Play in Three Acts, 34 VAND. L. REV. 265 (1981); Thomson, Some Ruminations on Rights, 19 ARIZ. L. REV. 45 (1977).


[FN41]. I intentionally avoid describing this view of precedential weight as ‘prima facie.’ That term, especially as used in the literature of rights and obligations, unfortunately suggests that a prima facie right, rule, obligation, or reason is not ‘real,’ or that it disappears when confronted with a more compelling reason. Such a view seems misleading, because nonabsolute rights, rules, obligations, or reasons still serve the purpose of requiring justifications for action stronger or different than would otherwise be required. Because rights, rules, obligations, and reasons always serve this purpose, some understanding is lost when we describe them as not real or as being, like the Cheshire cat, subject to disappearing instantly. See Schauer, Rights, supra note 40, at 228-30; Searle, Prima Facie Obligations, in PRACTICAL REASONING 81, 88-89 (J. Raz ed. 1978) (‘[There is] indeed a valid distinction, between what one has an obligation to do and what one ought to do all things considered. But this distinction is marked quite clearly in ordinary language and does not require the introduction of the term prima facie to mark it.’); see also Hintikka, Deontic Logic and its Philosophical Morals, in MODELS FOR MODALITIES (1969). Moreover, a reason need not be as strong as, for example, a presumption. If the force of a precedent were to be taken as presumptive, the question in the second case would be whether, having decided the first case as), it is possible in fairness to decide the second case as Z. See R. DWORKIN, supra note 6, at 113. As so characterized, those who would urge a different result in the second case carry the burden of proof. But precedent might even be said to constitute a reason (and therefore, ceteris paribus, a dispositive reason) without even having the force of shifting the burden of proof. For example, the question in the second case still might be simply one of what ought to be done, with the decision in the first case merely being one of the factors considered in this more open-ended inquiry.


[FN42]. My usage here varies from the standard sense in which a binding precedent is different from a precedent that might merely be persuasive. Thus, we say that the Supreme Judicial Court of Massachusetts is bound by a previous decision of that court in a way that it is not bound by a decision, even if ‘on point,’ of the Supreme Court of Idaho. Here the distinction between the binding and the nonbinding is largely one of jurisdictional ambit. And the question of jurisdictional ambit is in turn inextricably intertwined with the question of authority. Thus, decisions are not binding when there is no relationship of authority between one court or one decision and another.


[FN43]. See J. RAZ, THE AUTHORITY OF LAW 183-89 (1979); Simpson, The Ratio Decidendi of a Case and the Doctrine of Binding Precedent, in OXFORD ESSAYS IN JURISPRUDENCE 148 (A. Guest ed. 1961).


[FN44]. This was the traditional English rule, which has been somewhat relaxed of late. See generally R. CROSS, supra note 3, at 103-52.


[FN45]. See, e.g., Loving v. Virginia, 388 U.S. 1 (1967); Korematsu v. United States, 323 U.S. 214 (1944).


[FN46]. I. KANT, On a Supposed Right to Tell Lies from Altruistic Motives, in CRITIQUE OF PRACTICAL REASON AND OTHER WRITINGS IN MORAL PHILOSOPHY (T. K. Abbott trans. 1973), discussed in S. BOK, LYING: MORAL CHOICE IN PUBLIC AND PRIVATE LIFE 39-49 (1978); C. FRIED, RIGHT AND WRONG 54-78 (1978).


[FN47]. Or it is overruled. But the possibility of overruling is peripheral to the instant discussion. When a precedent is overruled it goes away, and it will not be an argumentative burden in future cases. When the government demonstrates a compelling interest in an equal protection case, for example, or when any right is overridden by particularly exigent circumstances, the right persists even though it may not control the outcome. See Gewirth, Are There Any Absolute Rights?, 31 PHIL. Q. 1 (1981); Schauer, Rights, supra note 40; Thomson, supra note 40. But when a precedent is overruled its force comes to an end, and thus it is misleading to analogize overruling a precedent to outweighing it or overridge it.


[FN48]. Recall Marsh v. Chambers, 463 U.S. 783 (1983), in which the Supreme Court justified its rejection of the 3-part test of previous establishment clause cases by saying that the historical pedigree of legislative chaplains distinguished this case from other establishment clause cases.


[FN49]. See, e.g., Goodhart, Precedent in English and Continental Law, 50 LAW Q. REV. 40, 56-58 (1934); Wade, The Concept of Legal Certainty: A Preliminary Skirmish, 4 MOD. L. REV. 183 (1940-1941). Useful deconstructions of the idea, partly consistent with what I argue here, include Westen, The Empty Idea of Equality, 95 HARV. L. REV. 537 (1982); Winston, On Treating Like Cases Alike, 62 CALIF. L. REV. 1 (1974).




[FN51]. See J. RAWLS, A. THEORY OF JUSTICE 136-42 (1973).


[FN52]. See generally A. KOCOUREK, AN INTRODUCTION TO THE SCIENCE OF LAW 165-85 (1930); Marsh, Principle and Discretion in the Judicial Process, 68 LAW Q. REV. 226 (1952) (background on the claims of uncertainty and nonpredictability).


[FN53]. The argument is often couched in terms of certainty rather than predictability. See R. WASSERSTROM, supra note 6, at 60-66. But some outcomes might be more predictable than others yet still be far from certain. Talk of ‘certainty’ establishes an unattainable standard that serves only to distract from the value of the ability to predict with reasonable confidence what the future might bring.


[FN54]. See H.L.A. HART, supra note 27, at 121-32.


[FN55]. Burnet v. Coronado Oil & Gas Co., 285 U.S. 393, 406 (1932) (Brandeis, J., dissenting); see also Sheddon v. Goodrich, 8 Ves. 481, 497, 32 Eng. Rep. 441, 447 (1803) (‘better that the law should be certain than that every judge should speculate upon improvements').


[FN56]. As Justice Brandeis argued with respect to constitutional adjudication in Burnet, 285 U.S. at 406.See also Smith v. Allwright, 321 U.S. 649 (1944).


[FN57]. See Kennedy, Legal Formality, 2 J. LEGAL STUD. 351 (1973).


[FN58]. See B. CARDOZO, THE NATURE OF THE JUDICIAL PROCESS 149-50 (1921); K. LLEWELLYN, supra note 5, at 64-65; R. WASSERSTROM, supra note 6, at 72-74; Heiner, Imperfect Decisions and the Law: On the Evolution of Legal Precedent and Rules, 15 J. LEGAL STUD. 227 (1986). The considerations that underlie a rule of precedent may also at times generate specific substantive rules. E.g., Epstein, A Theory of Strict Liability, 2 J. LEGAL STUD. 151 (1973).


[FN59]. See Shapiro, Toward a Theory of Stare Decisis, 1 J. LEGAL STUD. 125 (1972). It is worthwhile to note that this perspective could explain why the Supreme Court should be constrained by precedent, and, more cynically, why the Court purports to be constrained by precedent even when it is not.

On precedent in the particular context of Supreme Court adjudication and judicial review, the leading recent case is Akron v. Akron Center for Reproductive Health, 462 U.S. 416 (1983); see also Thornburgh v. American College of Obstetricians & Gynecologists, 106 S. Ct. 2169 (1986); Vasquez v. Hillery, 106 S. Ct. 617 (1986). For discussion of the issue, focusing on why the Supreme Court should or should not be less constrained by precedent than other courts, see Frickey, A Further Comment on Stare Decisis and the Overruling of National League of Cities, 2 CONST. COMMENTARIES 341 (1985); Frickey, Stare Decisis in Constitutional Cases: Reconsidering National League of Cities, 2 CONST. COMMENTARIES 123 (1985); Kelman, The Forked Path of Dissent, 1985 SUP. CT. REV. 227; Maltz, Some Thoughts on the Death of Stare Decisis in Constitutional Law, 1980 WIS. L. REV. 467; Monaghan, Taking Supreme Court Opinions Seriously, 39 MD. L. REV. 1 (1979); Stevens, The Life Span of a Judge-Made Rule, 58 N.Y.U. L. REV. 1 (1983).

[FN60]. I take the term ‘ruleness' from Kennedy, Form and Substance in Private Law Adjudication, 89 HARV. L. REV. 1685 (1976).


[FN61]. Think, for example, about Ronald Dworkin's attempt to relate judicial decisionmaking to questions about the institutional role to be played by courts in a liberal democratic society, R. DWORKIN, supra note 6, about Lon Fuller's conception of the internal morality of law, L. FULLER, THE MORALITY OF LAW (rev. ed. 1969), and even about H.L.A. Hart's search for a ‘minimum content of natural law’ that would attach to positivist theory the social goals of a political organization. H.L.A. HART, supra note 27, at 189-207.


[FN62]. I am referring here to equity as a philosophical and historical ideal, and not to the somewhat oxymoronic if more realistic notion of ‘rules of equity.’ For a useful discussion of the Aristotelian ideal to which I refer, see R. WASSERSTROM, supra note 6, at 84-117.


[FN63]. Senator Packwood recently stated, in explaining why he did not vote to convict on one of the articles of impeachment against Judge Harry Claiborne, that ‘I did not want to establish a precedent that, because you're convicted by a jury, that you're automatically impeached, or, if you're acquitted by a jury, you automatically can't be impeached.’ Det. Free Press, Oct. 10, 1986, at 12, col. 5.


[FN64]. See Koch, Confining Judicial Authority over Administrative Action, 49 MO. L. REV. 183, 210-20 (1984).


39 Stan. L. Rev. 571