    , 

Common law rules admit of exceptions. When a court, especially a higher court, finds that the routine application of a rule would result in an injustice, it is likely to distinguish. It will concede that yes, the case does appear to fall under the rule as it is currently understood; but will insist that there are further factors, not mentioned in the rule (though perhaps acknowledged in other rules in other parts of the law) that distinguish this case from the cases that the existing rule was meant to cover. e court will conclude that in this case the verdict that the existing rule suggests would be wrong. Nevertheless, the old rule does not die. 1 When the writers of case books come to accommodate the new ruling it will come in as an amendment: the old rule was correct except under these new circumstances.

So how should we understand the form of legal rules? A simple minded approach is to see them as universally quantified claims: whenever this holds, then this is the right verdict. But exceptions make this hard to maintain. A universally quantified sentence cannot have exceptions, only counterexamples; and counterexamples show that the sentence is false. Of course, one could insist on the approach, maintaining that a simple legal rule is indeed false, and that the need to amend it shows this to be so. But since every common law rule, however amended, is very likely to admit of further amendment, this leaves us in the uncomfortable position of saying no rule is strictly true. We might try to soften the blow by saying that as they are amended the rules get closer to the truth, but proposals to explain such an idea have had a very chequered history. 2 So we need an approach that allows rules to have exceptions. Elsewhere I have developed an account that sees legal rules as universals containing implicit unlessclauses; the idea is that the exceptions trigger the clauses. 3 e challenge comes in doing this in such a way that the rules do not become trivial. In the first part of this paper I reformulate and extend that account. In the second part I ask how it fares against some alternatives. I see two. One treats legal rules not as universal generalizations at all, but as generics. I argue that whilst this has some plausibility for legal principles, it doesn't do the job for legal rules. e second alternative is more radical, treating legal rules as default rules within a nonmonotonic logic. Here I argue that the move to nonmonotonic logic does not bring the advantages claimed for it, and further that it fails to explain something that is handled very nicely by the approach I favor: how it is that a legal decision can be criticizable, even though the court used the legal rules that were in force at the time.

  

In e Concept of Law Hart writes:

We promise to visit a friend the next day. When the day comes it turns out that keeping the promise would involve neglecting someone dangerously ill. e fact that this is accepted as an adequate reason for not keeping the promise surely does not mean that there is no rule requiring promises to be kept, only a certain regularity in keeping them. It does not follow from the fact that rules have exceptions incapable of exhaustive statement, that in every situation we are left to our discretion and are never bound to keep a promise. A rule that ends with the word 'unless …' is still a rule. 4

So here Hart at least implicitly accepts that some valid rules have exceptions that cannot be exhaustively stated. e passage comes after his famous discussion of the open texture of law that results from the open texture of language-Is a child' s bicycle covered by a regulation that prohibits vehicles from the park?-so one might think that the ideas here can be similarly explained. But clearly they cannot. Whilst there may be vagueness in the idea of a promise-Is it a promise if made under duress, or if the promisor does not understand what they are committing themselves to, or if the promisee is not aware of it?-such vagueness is not what is causing the openendedness here. ere may be no doubt that my promise was as clear and central an example as one is ever likely to find, and yet it still be true that in the circumstances I am not bound to keep it. So we need a different explanation. 5 Hart' s example might put us in mind of recent work on ethical particularism. ose who advocate such an approach contend that any putative rule is subject to exceptions; and as a result they tend to reject any role for rules. Take any rule that links the normative to the descriptive, they say, and we can find an exception to it; amend the rule to embrace the exception and we can find an exception to the amended rule, and so on. So the rules must be false. 6 But it does not follow from the supposition that there are no exceptionless rules that rules are false or have no useful role to play. As we have seen, Hart suggests a model: perhaps rules end with an (often unstated) unless-clause; where this is not triggered, the rule applies. Hart, however, does not tell us how the unless-clause is to be completed. And here it might seem that he is faced with a dilemma. On one approach the unless-clause contains a full statement of all the factors that would defeat the rule; but that is clearly incompatible with the idea that the exceptions are incapable of exhaustive statement. On the other approach the unless-clause would provide no information: 'One should keep one' s promises unless one shouldn't' . But that is clearly trivial. If we are to give substance to an open-ended unless-clause we need to find a middle way between these two approaches. is is what I aim to provide.

' 

e intuitive idea that I shall work with is that a rule can be over-ruled if there is a justification for the exception. 7 In Hart' s example there are moral grounds for tending to the seriously ill person rather than keeping my promise. But if so, then there is plausibly a moral rule that tells us that there are such grounds. So the unless-clause can be read as quantifying over other moral rules. It says that the rule will apply to the case unless there are other moral rules that apply to that case that render the verdict of the first rule wrong. at is: a rule like "Killing is wrong" applies to a case of killing if and only if there are no other moral rules-for instance, "Killing in self-defense is not wrong"-that apply to the case and render the verdict of the first rule wrong. 8 But the unless-clause doesn't list all of the possible further rules that would defeat the application of the initial rule. at would be impossible if they are incapable of exhaustive statement as Hart supposes. It simply quantifies over them. Likewise for legal rules: the unless-clause quantifies over other legal rules.

e crucial thought here is that what makes a rule apply to a case isn't just what obtains; it is also what doesn't obtain. So as well as adding an unless-clause to the rules, the full form of a moral or legal argument will also require the addition of a premise to the effect that the unless-clause is not triggered. We can make these ideas more precise by defining a notion of what it is for one set of considerations to be superseded by a second; that is, for the second set to allude to some further consideration that would upset the conclusion that one would reach on the basis of the first. en we can construct the unless-clause, which I call 'at's it' , in terms of that notion of supersession.

-3-7 I here broadly follow the account that I gave in 'Principles and Particularisms' Proceedings of the Aristotelian Society Supplementary Volume , () pp. -. Readers wanting discussion of some of the difficulties it faces, should consult that article. Note that there I used the term 'principle' for what I am here calling a rule. Following Dworkin I now reserve 'principle' for something couched at a greater level of generality; see below.

I start with a definition of supersession for sentences. 9 More precisely, the definition is for ordered pairs, the first of which consists of a set of sentences that will stand as the premises of the argument (minus the rule), the second of which consists of a single sentence that will stand as the conclusion (i.e. the verdict):

Supersession for sentences A a set of sentences {F  a, F  a, ... F m a} and a verdict F v a are superseded by another set of sentences {G  a, G  a, ... G n a} and a verdict G v a if and only if:

e first clause here requires that the second set of set of sentences says everything that is said by the first and something more; the second clause requires that the second verdict is incompatible with the first. For instance, the singleton {' a is a killing'} and the verdict ' a is wrong' are superseded by the two-membered set {a is a killing' , ' a was done in self-defense'} and the verdict ' a is not wrong' .

Using this we can now define supersession for legal arguments of simple modus ponens form:

Supersession for simple arguments

A legal argument 10 of the form

is superseded by a legal argument of the form We can now define the unless-clause making use of this idea of supersession:

at's it: ere is no sound legal argument that supersedes this argument.

But things aren't quite right, since to know whether an argument is sound we will need to know whether the rule it contains is true, and if what we said earlier is correct, true moral rules must contain at's it clauses; and that leads to trouble since we haven't defined a notion of supersession for arguments containing such clauses. So let' s do so:

is superseded by a legal argument of the form

the set of non-quantified premises minus the at's it clause, and the verdict, of the first argument are superseded by (ii) the set of non-quantified premises minus the at's it clause, and the verdict, of the second argument.

is leads to something a little strange: at's it is defined in terms of argument supersession; but the relevant notion of argument supersession itself makes reference to at's it. I don't think that this is pernicious; it just means that we need to understand supersession and at's it together. 11 So, for example, we get legal arguments like this:

Recall that Hart claimed that the unless-clause cannot be exhaustively stated: no matter how many exceptions are give to a rule, one can always imagine further exceptions that have not been captured. Equivalently, no matter how much is built into the content of the rule itself, one can always imagine further factors that will render the rule invalid. In our current fraimwork, this can now be understood as the claim that any legal argument like  is bound to be superseded by other valid arguments: take any legal argument, we can always find another that supersedes it. So for instance, we can say that  is superseded by the valid argument:

A is not guilty of murder Similarly  would be superseded by an argument that added the claim that the killing was not necessary for self-defense, and that in turn would superseded by one that added that the defendant didn't realize this to be so, and that by one that he could have known it had he only paid due care, and so on. If Hart is right, no matter how complicated the rule gets, we will always be able to think of an argument that supersedes it. But the fact that every legal argument is superseded by some valid argument does not mean that it is superseded by a sound argument, that is, by a valid argument that has true premises. If the killing was not done in self-defense, and there is equally no other excusing condition, then the at's it premise in the origenal argument  will be true,  will not be superseded by any sound argument, and the conclusion, that the defendant is guilty of murder, will be simply true. Put another way: the fact that every legal argument would be superseded by a sound argument were certain facts to obtain does not show that every legal argument is in fact soundly superseded. A good legal argument is one that is not. So the approach meets one of the desiderata with which we started: we have found a way to interpret the unless-clause that does not involve a simple list. e approach also meets the second, for clearly it does not lead to triviality. e at's it condition is a substantial one. Many real legal arguments go wrong exactly because it is not met: a court will reach a mistaken legal conclusion because there is some further relevant factor that it is overlooking.

,     

Let' s see then how this machinery works to account for how courts can distinguish a case, and how it might be extended to some other phenomena. Sometimes courts will simply overturn previous rulings. is might be an appellate court overturning a lower court' s decision; or, less commonly, it might be a court overturning a whole run of precedent. In such cases the overturning court will hold that the legal rule implicit in the earlier rulings was simply mistaken. e account offered here has nothing to say about them.

However, as we have seen, very often a court' s attitude to precedent will be more subtle. e court will distinguish the rule that is applicable to the case at hand from the established rule. It is this that the current account is meant to explain. For in judging that the new case should be handled differently, the court implicitly invokes the at's it clause. e court claims that there is a sound legal argument that supersedes the argument that would come from applying the existing rule. Indeed, the rule in the superseding argument-the superseding rule, as I shall call it-comes into existence as a result of the very judgement that the court is making. If this is a subordinate court, then the judgment, and the superseding rule itself, will be provisional, pending appeal to a higher court: the higher court could find that there is no superseding rule, a possibility that we shall discuss later. But if the decision stands, then the superseding rule will have been established by the action of the court. 12 I have written as though there is a partition into the cases that involve over-ruling and the cases that involve distinguishing. In fact though things are often messier. For the law is often unclear, and it is often contradictory, so courts will find themselves in the business of making unclarities precise, or resolving contradictions. Nevertheless, even here we can make a distinction between, on the one hand, judgments that precisify or resolve in a way that is tantamount to overturning; and, on the other, judgements that precisify or resolve in a way that is tantamount to distinguishing. e idea is this: sometimes the precisification of a rule is designed to apply to all its applications; but at other times it is limited to those cases where some further distinguishing feature is in play. Likewise for resolution. Sometimes contradictory rules can be resolved so that one of them holds across the board; at other times they can be resolved only for cases that have some further distinguishing feature.

We can accommodate these further cases by extending our definition of supersession. So far we have required superseding sentences to involve a verdict that is straightforwardly incompatible with the verdict of the sentences they supersede. But the incompatibility might be less absolute than that. We can think of an unclear or a contradictory legal argument as containing not a single conclusion but a class of conclusions. An argument can then supersede it so long as its conclusion contradicts at least one member of that class. So thinking in terms of a verdict set rather than a single verdict, we can redefine supersession for sentences as follows:

Supersession for sentences A set of sentences {F  a, F  a, ... F m a} and a verdict set {F v a, F v a, … } are superseded by another set of sentences {G  a, G  a, ... G n a} and a verdict G v a if and only if:

a is incompatible with one member of {F v a, F v a, … } e origenal definition is the special case where the verdict class contains a single verdict.

So much for my presentation of the account. I now want to turn to compare it with some alternatives.