I. EXAMPLES

Let me outline three types of cases where the principle of treating like cases alike is relevant. Consider, first, the case of a judicial decision that is based partly on an arbitrary choice. Determining the exact levels of punishment is a good case in point. Suppose, for example, that a previous legal decision, call it D1, determined that the punishment for an armed robber is ten years of imprisonment. Let us assume that there are no particular reasons that would support imprisonment for ten years as compared to, for example, nine or eleven.⁴

Now assume that a very similar armed robbery comes before a new court (that is, assume that the cases are identical in all relevant respects, whatever you take the relevant respects to be), say D2. To suggest that there is a principle of treating like cases alike, amounts to the suggestion that even if there are no particular reasons to prefer ten years over eleven (or nine), the court in D2 ought to follow the decision in D1 and impose the ten years. And this would suggest that it would be wrong for the court in D2 to opt for a different level of punishment of, say, eleven years.

It is tempting to say that none of this should matter. After all, the assumption was that within the range of nine to eleven years, any decision would be just as good as any other. So why should we care? Well, for one, the convict would care, and quite a bit. Could he not reasonably complain that if the convict in D1 was punished with ten years of imprisonment, it would be wrong to put him away for eleven years? (Recall that we have assumed that the cases are identical in all relevant respects.) True, the convict in D2 cannot claim that the eleven years would be in itself somehow disproportionate or inappropriate. We have assumed the contrary. But can he claim that, given the previous decision in D1, it would be wrong or unfair? And if the convict is right about this, should not we care about it too?

Now consider the case of a judicial decision that is based on a choice between incommensurable values. Suppose that the court in case D1 had faced a choice between two decisions, A and B, and that the difference between these options is as follows: Option A would score higher on promoting value P (say, economic efficiency) and somewhat lower on promoting a competing value Q (say, environmental protection), whereas option B would amount to the opposite; B would score higher on promoting value Q and somewhat lower on promoting value P. Let us assume, further, that the comparison between A (P>Q) and B (Q>P) is basically incommensurable or, in any case, not determinable by reasons.⁵

Now suppose that the court in D1 opted for A. Should this choice amount to a binding decision, so that in a future case, D2, which is similar in all relevant respects, the court ought to follow D1 only because like cases should be treated alike? Or would it be warranted on the part of the court in D2 to recognize that the choice in D1 was not supported by reasons that can determine a preference of A over B, and therefore the court in D2 should be free to opt for decision B?

Finally, consider the case of moral vagueness. Judicial decisions often have to determine whether to apply a certain moral (or other evaluative) concept to the case at hand.⁶

The case may turn on the question of whether, say, action P is really “dishonest,” or not. Honesty, let us assume, is a moral concept, and like any other concept, it would have some borderline cases due to its inevitable vagueness. A deliberate lie, for instance, is typically dishonest, but not always. Some “white lies” are not dishonest at all, and some would be borderline cases. (Consider this example: Teachers and educators may encourage a troubled student by slightly exaggerating his or her achievements or talents, and so on; surely, at some point, the exaggeration would become dishonest, but it is unidentifiable at what point such “white lies” become dishonest.) In other words, there is no reason to assume that moral concepts are free from vagueness.⁷

And there is no reason to deny that some of the decisions courts have to make are about the application of such concepts to borderline cases. But then, if a case is a morally borderline one, the decision of the court cannot be described as one that is determined by the relevant moral reasons. (Perhaps the decision is warranted by other relevant moral reasons, but that is beside the point now.) If the court has to decide whether action P is “dishonest,” and P falls within the borderline of “honesty,” then the decision is, ex hypothesis, underdetermined by what “honesty” really requires.

You might think that such legal decisions on morally borderline cases should be avoided. Perhaps so, but they cannot be avoided, not entirely anyway. It is true that in some cases the law purports to have a margin of safety, phasing out borderline cases as if they are “no” cases. A good example is the rule of leniency in criminal law, prescribing that unless an action falls well within the meaning of the offense, it should be regarded as if it does not. This is a noble attempt to phase out borderline cases, but it does not always work. First, these kind of precautions are rare outside the sphere of the criminal law, and even there they are not (and perhaps cannot be) consistently applied. Second, and more important, as we know very well, vagueness cannot be eliminated, and even if you employ a generous margin of safety, you are still left with second-order vagueness (viz., vagueness about where the borderline cases begin).⁸

So now suppose that the court in D1 decides that action P, which, we assumed, is a borderline case of “honesty,” is dishonest. Should this decision be legally binding on future courts? (I am not asking whether it is a legal duty, but whether it ought to be one.) If the court in D2 faces a very similar case (that is, again, assume that D2 is identical in all relevant respects), should it be bound to decide D2 according to D1 just because the cases are similar, or should it be allowed to reach the opposite conclusion?

These three types of cases are not meant to be exhaustive. Other cases would have a similar structure. What they all have in common, however, is the fact that the reasons that would justify the previous ruling of the court do not determine the decision in the new case, even when the new case is basically identical in all the relevant respects.⁹

Thus the question is whether the mere similarity of the new case to the previous ones is sufficient to consider the previous rulings legally binding; should the new case be decided alike just because it is alike, even if it is underdetermined by the reasons justifying the previous rulings?

II. JUSTICE SHOULD BE SEEN

There are two main principles that could justify an affirmative answer to this question. Let us consider, first, the idea that is (loosely) expressed by the saying that “justice should not only be done, but should manifestly and undoubtedly be seen to be done.{”}¹⁰

Admittedly, this is a rather crude formulation, since we have seen in detail that there is no real issue of justice involved here. But there may be a value to uniformity of legal decisions in being seen as uniform by the parties concerned, even if there is nothing else to commend the decision but for its similarity to previous decisions. This value—call it the need for justice to be seen, if you like—consists of the assurance that litigants may gain by coming to realize that they are treated equally with others in similar situations. The more you can come to recognize that you are being treated exactly like others, the less you have reasons to suspect bias, prejudice, and other distortions that may have affected judicial decisions.

This is a complex value that has both an epistemic and a symbolic aspect. The epistemic aspect relates to the value of uniformity as a safeguard against potential bias. Consider, for example, the case of our armed robber who faces punishment. If he knows that the judge has the discretion to disregard the previous ruling that set the punishment at ten years, he would not be unreasonable to suspect that if he receives eleven years of imprisonment, it may have been a result of a particular prejudice against him. After all, as we have assumed, there is no particular reason that compels this decision, and therefore the suspicion of bias or prejudice might loom large. Dispelling such suspicions, even when they are actually unfounded, may be valuable in itself. Furthermore, there is a symbolic value to uniformity of decisions that is not insignificant. By requiring the courts to treat like cases alike, the law purports to assure the litigants that they are treated equally by the law. Even if this is actually an equality that is not called for by the relevant reasons that should determine the case, a public affirmation of such equality may have a symbolic value that is desirable in itself.¹¹

The main problem with this reasoning is that it provides some support for the principle of treating like cases alike, but not a very strong one. Generally speaking, concerns about appearances are typically weak and easily defeasible. Consider, for example, the concerns about the possibility of judicial bias or prejudice. One way to alleviate the worry that we may have about the possibility of bias on part of judges is to try to reduce their discretion as much as possible. But this can be a very costly way of achieving what we want. After all, we need judges partly because we expect them to exercise discretion. There may be other ways of reducing the suspicion of bias and prejudice that are just as good or perhaps even better. The requirement from judges to provide detailed justifications of their rulings that are rendered public, the possibility of appeals, and other forms of institutional checks and balances may constitute more effective ways of convincing potential litigants that they have not been discriminated against.

You may think that this is just a rough generalization, and of course it is. Furthermore, there may be certain situations in which the danger of a particular bias or prejudice looms large, and then perhaps in those cases, no precaution would be too much precaution. (Unfortunately, certain situations in which racial prejudices are all too present are a notable example that comes to mind.) In other words, it would be fair to say that in some cases the value of justice being seen and not just done is more important than in other cases.¹²

Nevertheless, and as a rough generalization indeed, I would venture to guess that public scrutiny and institutional constraints would work better than a practice that would require judges to “treat like cases alike.” After all, if the judge is indeed biased or prejudiced, it would not take a great deal of ingenuity on her part to show that the case in front of her is not really like the other cases. Judicial resourcefulness in distinguishing cases is almost unlimited.¹³

III. PROTECTED EXPECTATIONS AND THE VALUE OF PREDICTABILITY

Perhaps a much stronger case for the principle of treating like cases alike can be made by reference to the value of protected expectations. Potential litigants may have relied on previous judicial rulings and formed expectations about the law that it would be wrong to frustrate. But we must proceed with caution here: Not every expectation potential litigants may have about the law is one that it would be wrong to frustrate. The expectation must be reasonable and legitimate. Part of what makes expectations about the law reasonable and legitimate is that the relevant type of expectations is supported by legal practice. To take an extreme example, in a legal system that does not recognize a doctrine of precedent, potential litigants cannot rely on judicial decisions to make legitimate predictions about the law. If the law tells you that the decisions of higher courts are not binding, you cannot have a legitimate expectation that they would be. So the question of what expectations about the law are legitimate is partly, but crucially, a matter of legal practice. This in itself makes the argument about treating like cases alike somewhat circular: Unless there is a judicial practice of treating like cases alike, how can you claim that you have had a legitimate expectation that such cases would be treated alike?

This is not so simple, however. Sometimes we want to make an argument about the importance of protecting expectations even if the type of expectations in question is not yet supported by legal practice. In such cases we argue that legal practice ought to protect the relevant expectations because that would be a good thing to do. Certain social goals are better achieved if the judiciary follows bright-line rules that parties can rely on and thus predict, with greater accuracy, the content of the law that the courts will enforce. This is the kind of argument that is often made on behalf of clear canons of statutory interpretation: The more such canons of interpretation are clearly laid out and consistently followed by the courts, the greater stability and predictability the law would have, and this, it is argued, is generally a good thing.¹⁴

So can we make a parallel argument about the principle of treating like cases alike? Can we say that even if it is not a practice-based expectation that needs to be protected here, it is the kind of expectation that should be positively encouraged by the courts?

It seems to me that such an argument would be more cogent in some cases than others. For instance, it is rather doubtful that accurate predictability is valuable in the case of punishments. There is no reason to assume that potential criminals should be able to predict as accurately as possible what their punishment might be. It would probably not enhance the deterrent effect of the penal system,¹⁵

and as long as punishment is kept well within permissible boundaries, fairness should not be a concern either. Criminal punishment is not like a tax that criminals have to pay on their antisocial behavior. Taxes should be predictable indeed, because people should be allowed to calculate the costs of their economic activities ex ante and make their relevant choices. In contrast to tax law, the criminal law is there to prescribe modes of conduct that are not permissible. It is not the general purpose of the penal system to enable the potential criminals to calculate in advance whether they would rather follow the law and not be punished or violate it and pay a fixed penalty for it. To be sure, I am not suggesting that we have no reason to maintain a principle of proportionality in punishment. Far from it. My only suggestion is that within the range of punishments that are fair and proportional to the crime, accurate predictability serves no desirable social goals.

It may be tempting to think that the case of punishment is unique. In most other areas of the law, it may be argued, the greater the uniformity and predictability of judicial decisions, the better. On closer scrutiny, however, it should become clear that even as a rough generalization, this is not true. Increasing the predictability of judicial decisions is never costless, and often the costs will outweigh the benefits. The more judicial decisions are predictable, the more rigid they are. Generally speaking, rigidity is undesirable because it tends to impede judicial experimentation, innovation, and other potentially desirable social changes. It is a truism, but nonetheless an important one, that there is always a delicate balance that needs to be maintained between the values of predictability of judicial decisions and the values that call for their flexibility.

Consider, for example, the case of judicial decisions that reflect a choice between incommensurable values. Remember the example: We have assumed that the court in D1 faced a choice between option A, which scores higher on value P and lower on a competing value Q, and option B, which is the opposite, scoring higher on Q and lower on P. We have also assumed that the choice between A and B is basically incommensurable. Now the question is whether, faced with a similar case, the court in D2 should follow the decision in D1 only because the cases are similar and a decision in D1 had already been rendered, say, to prefer A over B. Admittedly, an affirmative answer could be justified by an appeal to the ideal of predictability of judicial decisions. But this can hardly settle the issue. There is something to be said in favor of the opposite answer, allowing the court in D2 to experiment with a different choice of values that is, ex hypothesis, not necessarily inferior to the choice of the court in D1.

But why should we allow the courts, you may ask, to engage in social experimentations? For two main reasons. First, there is an epistemic issue here. The actual effects and ramifications of some judicial decisions are often very difficult to assess in advance. It may take years, if not decades, to realize and adequately appraise the social consequences of some judicial decisions. Experimenting with different decisions under similar circumstances might facilitate a better assessment of the effects of various evaluative and other social choices that the courts make.¹⁶

Second, and perhaps more important, there is an issue of pluralism and diversity that might be involved here. Even if certain choices between values or mixed goods are, objectively speaking, incommensurable (or otherwise indeterminate by reasons), different people might nevertheless have a preference for one over the other option that is not irrational. In other words, incommensurability does not entail indifference. (Even if we assume, for example, that the choice between the life of an academic and that of a merchant banker is basically incommensurable, it does not mean that my choice of the former is arbitrary or something that I should not deeply care about.)¹⁷

Thus, allowing the courts to make different choices under similar circumstances may be justified by the need to cater to different populations with divergent needs and preferences. As long as the relevant judicial decisions are morally and otherwise within the bounds of permissibility (as we have assumed all along), courts should be allowed, at least in certain cases, to reach different decisions that reflect different preferences amongst otherwise incommensurable values. The uniformity of decisions that is dictated by the principle of treating like cases alike seriously undermines the courts' ability to respond to the needs of diversity and pluralism.

To be sure, I am not suggesting that these reasons for forgoing uniformity are necessarily present in each and every case. Far from it. But often one or both of them are, and this should count against the assumption that the value of predictability of judicial decisions ought to prevail. Thus, treating like cases alike is not necessarily desirable or even warranted.

IV. THE ROLE OF ANALOGY AND COHERENCE RECONSIDERED

So far I have considered the principle of treating like cases alike only in those cases in which the ruling of a previous court is underdetermined by reasons. It may be worth exploring another possibility: The principle of treating like cases alike may be thought to apply to cases that are determined by similar reasons. In order to explain what I have in mind here, let me draw a distinction between two forms of analogical reasoning in adjudication. Let us call it strong and weak analogy.

In the case of strong analogy, the court draws on previous rulings in order to extract the underlying justification for those rulings and seeks to apply that justification to the novel case. Suppose, for example, that a court decided in D1 that automobile manufacturers must fully disclose the exact technical specifications of the safety features of the cars they manufacture. Now suppose that the new case, D2, concerns a requirement to disclose safety features of mechanical toys. A reasoning by analogy would run as follows: The court in D2 must first try to surmise what the underlying reason or principle is that actually justifies the court's ruling in D1 to require full disclosure in case of automobile manufacturers; suppose it is principle P. Now, having realized that P justifies the ruling in D1, the court should decide whether P also justifies a similar ruling in D2, with respect to mechanical toys. If it does, the same principle is applied by analogy to the new case. Thus the effect of reasoning by strong analogy is that a certain principle or reason previously relied on by the courts is extended to novel cases, but this is only an extension of application, not of the underlying principle itself.¹⁸

By contrast, reasoning by weak analogy does not apply the very same reason or principle that justifies the previous rulings. Instead, roughly, we would want to say that the court relies on a similar principle, one that would be in accord with it even though it is not the same. Thus consider the following cases: In D1, the court decided that liability-exemption clauses in standard-form contracts should be subject to a requirement of “reasonableness.” Suppose that the underlying reason for this ruling (call it P1) consists in the concern about the disparity of information and bargaining power in standard-form contracts between the two types of parties to such contracts. Now suppose that the new case, D2, concerns a question about the validity of liquidated damages of standard-form contracts, and the question is whether any upper limit should be imposed on such liquidated damages. Presumably the underlying justification for the ruling in D1, that is, P1, does not by itself apply to the new question of limits on liquidated damages. After all, different considerations may apply to the questions about exemption clauses and remedies. But there is a clear sense in which a similar principle might be at work in D2, one that would be in accord with the justification of D1; namely, a concern about disparity of power between contractual parties to standard-form contracts and the need to protect the relatively weak and vulnerable side. So there is a sense in which a decision by the court in D2 to impose some limits on liquidated damages of standard-form contracts would be justified by a principle, say P2, that is similar to P1 that justifies the ruling in D1.

Similarity is a very vague concept, of course. But we can make it more precise: We could claim that given P1, P2 is required, or at least justified, by considerations of coherence. The assumption would have to be that P2 coheres better with P1 than its alternative. Therefore it would also make sense to say that given P1, P2 would be legally justified.

You may think that there is a different type of reasoning going on here: Perhaps the court in D2 simply takes the principle underlying the previous ruling, P1, to a higher level of abstraction and then applies the more abstract principle to the case at hand. Perhaps this is so, but then you have to keep in mind that the higher we ascend in levels of abstraction, the less determinate the principles become. In any case, I suspect that this description does not undermine my point. Once you try to justify the application of the more abstract principle to the new case, the same types of concerns will be present: The court in D2 would have to justify its assumption that the more abstract principle has sufficient legal support, as it were, and ex hypothesis, the support for the abstract principle and its legal relevance can come only from its similarity, or coherence, with P1 (and other similar previous decisions).

To sum up, the use of strong analogy in adjudication is not an instance of treating like cases alike because it actually amounts to the application of basically the same reasons to novel cases. In contrast, it is arguable that the use of weak analogy can be rationalized as an instance of treating like cases alike and one that is justified by a direct appeal to the value of coherence in adjudication.¹⁹

Notably, however, once we invoke the concept of coherence here, it should become clear that it takes more than one or two previous rulings to justify reasoning by weak analogy. There must be a larger set of previous decisions pulling in a certain direction, as it were, to make sense of a coherence justification of new decisions. That is so because any pair of noncontradictory propositions can be made to form part of a coherent set under some set of other propositions. Unless there is a relatively large set of previous rulings that instantiate a significant number of related reasons or principles, a coherence justification can hardly yield any specific results.

Be this as it may, the value of coherence, I suggest, constitutes both the rationale and the limits of applying the principle of treating like cases alike in the context of weak analogy. The rationale must be derived from the value of coherence in adjudication, and the limits of this value would also point to the limits of weak analogy.

A detailed account of the value of coherence in adjudication would exceed the scope of this article.²⁰

I will confine myself to two concluding remarks. First, I want to show that, generally speaking, coherence is not the underlying justification for the principle of treating like cases alike, with the notable exception of weak analogy. Second, as I hope is clear by now, it should not be assumed that weak analogy is necessarily justified whenever an occasion for its use presents itself. As in other cases of treating like cases alike, there may be very good reasons to treat like cases differently. All this, I admit, is very inconclusive. But sometimes the conclusion of a normative analysis is that there should not be any conclusive answer to the question dealt with. The principle of treating like cases alike is one of those issues.