II. CONTROL

A natural first thought is that a norm's membership in a legal system depends on who can control that norm. Matthew Kramer says:

In my terminology, Kramer's test says: a norm adopted by some legal system is subject to the control of people external to that system. A norm local to that system is not.

Kramer's test deals well with certain examples. The norm that permits roulette is not part of English law because it is subject to the control of Monte Carlo officials. By contrast, the norm that says that a gambling debt is recoverable only if it is incurred to play a lawful game is part of English law because it is not subject to the control of anyone outside of the English legal system.

While there is much to like about Kramer's test, it suffers from a number of problems. Suppose that you and I agree that any dispute arising under our contract will be resolved according to South African customary law. Customary law, Kramer says, “is not really within anyone's effective control.”Footnote ⁶ So, South African customary law is not really under anyone's control, including the control of anyone external to the English legal system. Kramer seems to be committed to the view that if an English court relies on South African customary law to resolve our dispute, it becomes part of English law. But it does not.

Similarly, judges often use semantic and grammatical norms to help decide cases. These norms are not under anyone's control; they are “free-floating,”Footnote ⁷ to use Kramer's term. When these norms are legally relevant, it would seem that Kramer's test wrongly counts them as local laws. Now, Kramer recognizes that judges rely on semantic and grammatical norms. But he denies that they “serve as justificatory bases for official decisions.”Footnote ⁸ He says:

I agree that judges do not treat semantic and grammatical norms as an independent basis for legal decisions, a point I will return to in Section VIII. But it is false that judges do not treat such norms as reasons for reaching one decision over another.

Consider Smith v. United States.Footnote ¹⁰ The United States Criminal Code imposed a criminal penalty on whoever “during and in relation to any crime of violence or drug trafficking crime … uses … a firearm.” The issue for the United States Supreme Court was whether the penalty applied to the “use” of a MAC-10 automatic weapon in trade for cocaine. To help resolve this issue, the court consulted the dictionary definition of “to use.” Partly because the accused's “use” of the MAC-10 “fell squarely”Footnote ¹¹ within this definition, the court found that the penalty applied. Here, the semantic norms reflected in the dictionary definition of “to use” provided a crucial premise in the reasoning that led to an overall conclusion about the accused's liability. The semantic norms helped to justify that conclusion, in other words. And yet these norms did not thereby become part of American law.Footnote ¹²

So far, I have been pointing to adopted norms that are not under anyone's control.Footnote ¹³ In addition, there are adopted norms that are under local control. For example, many administrative policies, which are not laws, are under the control of government ministers, who wield lawmaking power. There is also reason to think that Kramer's test is underinclusive, because some norms that are under external control are also part of local law. For instance, some legal systems incorporate international treaty law into local law. And yet treaty law may be under the control of external actors (e.g., other states, international organizations, international courts). So, despite its attractions, Kramer's test must be rejected.

III. CREATION

Scott Shapiro flips Kramer's idea on its head. According to Shapiro, what matters is not the lack of control that external institutions have over a norm, but a specific type of control local institutions have over them. Shapiro says that “for two enacted rules to be part of the same system they must have been created according to the power-conferring provisions of the same shared plan,” where each legal system has a shared “master plan.”Footnote ¹⁴ Thus, I take it that Shapiro would distinguish local and adopted norms as follows: whereas norms local to a legal system are created under powers conferred within that system, norms adopted by it are not.

Monte Carlo law is not created under the power-conferring norms of English law. Neither is South African customary law or the norms of grammar. Shapiro's test correctly says that none of these norms are part of English law. By contrast, the norm that a gambling debt is recoverable only if it is incurred to play a lawful game was created by English judges and legislators using powers conferred within English law. Shapiro's test correctly says that the debt-recovery norm is part of English law.

Although Shapiro's test is initially plausible, I think it is pretty clear that it yields unacceptable results. Consider the norms of English customary law. Customary norms emerge through the accretion of certain attitudes and actions in a group.Footnote ¹⁵ They are not created through the exercise of a power conferred by any norm. Hence, they cannot be created under powers conferred by the norms of any legal system. Hence, according to Shapiro's test, English customary law cannot be part of English law. To be clear, I grant that customary norms are part of a legal system in virtue of norms of that system. We could say that customary norms are laws of a legal system in virtue of its “master plan,” or in virtue of recognition under certain of its rules. What we must not say—and this is all that matters for assessing Shapiro's test—is that customary norms are themselves created under powers conferred within the legal system. They are not.Footnote ¹⁶

Let us suppose, for the sake of argument, that customary norms are created under powers conferred within a legal system. It would follow that international customary law is created under powers conferred within a legal system. Which system? The answer is certainly not: the Israel legal system. We know that because many norms of international customary law predate the Israel legal system, and a norm cannot arise in a legal system that does not yet exist. According to Shapiro's test, this means that international customary law cannot be part of the Israel legal system. On the contrary, though, “[a]s in most states, in Israel, too, international custom is automatically part of municipal law.”Footnote ¹⁷ It is “part of the law of the land.”Footnote ¹⁸

Finally, consider the relationship between European Union law and United Kingdom law. Certain types of EU law are given effect in the UK through a domestic statute, the European Communities Act 1972.Footnote ¹⁹ The key section is 2(1), which says:

As Section 2(1) acknowledges, EU law is “created … by or under the Treaties.” It is not created under powers conferred within UK law.Footnote ²⁰ According to Shapiro's test, it follows that EU law cannot be part of UK law. But this is at odds with how those who work within the law think. They think that EU law is part of UK law. The European Court of Justice, for example, takes the view that EU law is “an integral part of the legal system of the member states” and “forms part of the law of those states.”Footnote ²¹ The Supreme Court of the UK agrees. “EU law [is] a source of UK law,” it says, and forms “part of its domestic law.”Footnote ²² So, not all laws of a legal system are created under powers conferred within that system. Shapiro's test must also be rejected.

IV. SUPPORT

Kramer and Shapiro both focus on who has the power to make or change the norm at issue. By contrast, for Joseph Raz, “[t]he reasons for enforcing a norm, and the attitude of the courts and the legislature to its enforcement, are the crucial factors.”Footnote ²³ Raz says:

Simplifying a bit, the idea is that adopted norms are either (1) norms of another system recognized out of respect for the way that some community regulates its own affairs; or (2) norms created under local powers designed to help people manage their affairs as they like. Local norms meet neither condition. The first half of the test is meant to exclude foreign laws. The second half excludes “contracts, the regulations of commercial companies, and the like.”Footnote ²⁵ What is supposed to unite these conditions is the idea that one of the law's functions is to “support other social arrangements and groups.”Footnote ²⁶

Raz's test is both underinclusive and overinclusive, however. It is underinclusive because it counts as adopted at least some norms of provinces and states in federal systems. For example, the law of Canada grants Quebec the power to make laws with respect to education, partly to enable Quebec to arrange its affairs as it desires. These laws are made in some sense with the “consent” of those subject to those laws, i.e., the people of Quebec. According to (2), Quebec's education laws are not part of the Canadian legal system. But, as Keith Culver and Michael Giudice note, Quebec's laws “are thought of, spoken of, and function as norms of law in Canada, and would be very oddly characterized … only as norms capable of being adopted by a federal system.”Footnote ²⁷

Raz's test is overinclusive because some adopted norms are applied by judges for reasons that have nothing to do with supporting other social arrangements and groups. In R v. North and East Devon Health Authority, ex parte Coughlan,Footnote ²⁸ for example, a health authority promised Coughlan that she would have a home for life in a new nursing facility if she agreed to move from the hospital where she had been living for the previous twenty years. Coughlan agreed. After she had moved, the health authority tried to renege on its promise. Coughlan complained and the Court of Appeal for England and Wales agreed with her: by trying to break its promise, the health authority had acted unlawfully.

Coughlan is a seminal case in the development of what in English law is known as the “doctrine of legitimate expectations.” Roughly, that doctrine says that a public body has a legal duty to keep its promises. Under the doctrine of legitimate expectations, English courts regularly enforce promises by public bodies. These promises are not created through the exercise of a legal power. Nor are they enforced out of “respect” for the way that administrators seek to regulate their activities. Rather, judges enforce these promises because they think it is unfair for the government to say one thing and do another, at least if a person relies on what the government has said, as Coughlan did.Footnote ²⁹ In such cases, a norm (the promise) is legally relevant, not excluded by either condition in Raz's test, and yet is not a law.

So concludes my survey of the existing answers to the boundary question. It has not, I am afraid, been a very fruitful survey. Each answer is vulnerable to obvious counterexamples. Later, I will show why there is a kernel of truth in each of the existing answers. For now, it is time to explore a new way of thinking of the boundary question.

V. A DOCTRINAL PERSPECTIVE

Thus far, I have focused on how legal philosophers have approached the boundary question. But they are not the only ones to think about law's boundaries. Doctrinal scholars think about law's boundaries, too. Of doctrinal scholars, I think that constitutional scholars have come closest to the truth.

In Commonwealth jurisdictions, a “constitutional convention” is a term of art for a nonlegal norm of a constitution. Conventions play a central role in many Commonwealth constitutions. For example, there is a convention of the British constitution that the monarch will grant royal assent to bills passed by Parliament. Were the Queen to refuse assent to such a bill, she would act unconstitutionally but not—and this is the point—unlawfully. Because of their practical importance, constitutional scholars have thought a lot about the nature of constitutional conventions. In particular, they have wondered: What is the difference between a law and a convention? The standard answer, originating with AV Dicey, is that conventions are not laws because courts “recognize” or “enforce” laws, whereas they do not recognize or enforce conventions.Footnote ³⁰

To this standard answer there is a standard criticism: sometimes, courts do recognize and enforce conventions.Footnote ³¹ Dicey's critics have in mind cases like Copyright Owners Reproduction Society Ltd v. EMI.Footnote ³² In the 1920s, the Imperial Parliament at Westminster had the legal power to legislate for Australia. However, there was a constitutional convention according to which the Imperial Parliament would not legislate for Australia without its consent. The question in this case was whether a 1928 copyright statute of the Imperial Parliament extended to Australia, in which event the defendant record companies would have to pay higher royalties than they would otherwise. The High Court of Australia reasoned to a conclusion this way: in light of the consent convention, if Australia did not consent to a statute, Parliament would be presumed not to intend the statute to apply to Australia. Because Australia did not consent to the 1928 statute, Parliament should be presumed not to intend the 1928 statute to apply to Australia. Nothing in the statute rebutted that presumption. Therefore, the statute did not apply to Australia.

Cases like Copyright Owners seem to show, contra Dicey, that conventions have legal consequences. However, just as Dicey has critics, he has defenders. These defenders admit that judges give legal effect to conventions. But they maintain that judges do not give conventions legal effect in the same way as they do laws. The British constitutional scholar Geoffrey Marshall writes that in cases like Copyright Owners:

And a little later Marshall concludes:

Marshall's idea, I take it, is that whereas laws are “directly” applied or enforced, conventions are merely “indirectly”Footnote ³⁵ applied or enforced. There is also the suggestion that conventions are legally significant through or by virtue of their influence on laws.Footnote ³⁶

The quotes above are vague and the claims made in them are not fully developed. Neither Marshall nor Dicey's other defenders test the distinction between direct and indirect relevance against examples beyond the constitutional sphere. Still, Marshall was on to something important. The difference between direct and indirect relevance is, I think, the key to distinguishing local and adopted norms. In the sections following, I state the distinction between the two types of relevance precisely, then apply the distinction to the two types of norms.

VI. DIRECT AND INDIRECT RELEVANCE

To see the difference between direct and indirect relevance clearly, suppose that a norm of English law says that a deceased person's property goes to the person's next of kin if he or she dies intestate. If this norm applies to a case, then the norm establishes the truth of a conclusion about rights and duties as a matter of English law, namely, that the next of kin has a right to the deceased's property. That conclusion may be defeasible: the right may not be absolute. It may be that the norm is not essential for the conclusion: perhaps the next of kin would have been entitled to the property anyway, by some other route. But, if the norm applies, then on its own it makes true a conclusion about people's rights and duties within English law.Footnote ³⁷ The norm leads to that conclusion directly, we can say.

Not everything of legal relevance is of direct relevance. Suppose the following are facts in a case:

• that the makers of the intestacy norm intended it to mean such-and-such

• that so-and-so died intestate

• that so-and-so is related to the deceased by such-and-such degree of consanguinity

Of these facts, the first is (at least on certain theories of interpretation) relevant to the content of the intestacy norm. The second is relevant to the applicability of that norm in a case. The third is relevant to the exact consequences of the norm in a case, i.e., to working out who exactly is the next of kin and thus who exactly has a right to inherit. Each fact helps in deciding on people's rights and duties. Each “figures in the truth conditions of propositions of law,”Footnote ³⁸ to use Dworkin's phrase. But the legal relevance of each is mediated or indirect—it occurs through or via the fact's impact on the intestacy norm.

It is not only facts that are indirectly relevant. Some norms play a similar role: they are relevant to deciding on people's rights and duties because they are relevant to deciding on the content, applicability, or exact consequences of some other norm. I stipulate the following definitions, for a legal system L:

Two clarifications are in order. First, under these definitions, a norm can be of both direct and indirect relevance. That is, a norm can be relevant in its own right (direct relevance) and relevant to the meaning or operation of a norm that is directly relevant (indirect relevance). Second, there is no suggestion that a norm's content, applicability, and consequences can be sharply differentiated. One can easily blur into the other.

VII. THE RELEVANCE OF LOCAL NORMS

In all of my examples, the local norms are directly relevant. If the debt-recovery norm applies in Saxby, then it follows directly that Saxby has a right to recover the debts. If the criminal firearms penalty applies in Smith, then it follows directly that Smith is liable to a term in prison. If the doctrine of legitimate expectations applies in Coughlan, then, without more, we can infer that the health authority has a legal duty to provide Coughlan with a home. And so on. The norms in these examples do not always apply; but when they apply, in combination with the facts of the case, they suffice to yield certain conclusions within local law.Footnote ³⁹

Are these examples generalizable? That is, is every local norm directly relevant? The question requires some careful handling. Suppose that a local law says that you are entitled to compensation for personal injuries suffered while driving, provided you were driving in a lawful manner at the time. To decide on whether the compensation norm applies, we need to decide on whether you complied with various road traffic norms (e.g., speed limits). So, the road traffic norms bear on the applicability of the compensation norm. What this shows is that a local norm may be both indirectly and directly relevant. While the road traffic laws are indirectly relevant when deciding on compensation, they are directly relevant to deciding on a range of other rights and duties (e.g., the speed at which you are legally permitted to drive).

It is true, however, that legal systems do contain laws that are of mere indirect relevance. Here is Section 6 of Britain's Interpretation Act 1978:

This provision is not directly relevant: it does not, on its own, establish the truth of any conclusion about people's rights and duties. Rather, the section's legal effects arise indirectly, through or via norms that use either the masculine or the feminine but not both, or the singular or plural but not both. There are many similar examples: other legal definitions, limits on the territorial or temporal applicability of norms, laws that establish legal categories and statuses, and so on.

These examples show that some laws are merely indirectly relevant. But they do not show that some legal norms are merely indirectly relevant. To explain why, I want to borrow Raz's distinction between norms and rules.Footnote ⁴⁰ Although some norms are rules, some norms are not (an order, say, is a norm but not a rule, because it is specific not general). More importantly for my purposes, although some rules are norms, some are not. As Raz says:

Rules that are not norms can be “explained by explaining their logical relations to [rules] which are norms.” “Their whole point and function,” Raz says, “is exhausted by their impact” on the rules that are norms. Section 6 of the Interpretation Act 1978 is a good example. It does not impose a duty, confer a power, or grant a permission. It does guide conduct, but “indirectly,” through its impact on the interpretation of other rules. So, Section 6 lays down a law that is a rule but not a norm. In general, laws that establish definitions, territorial limits, statuses, categories, and so on are rules but not norms.

Let us reserve the term “rule” for a rule that is not a norm and “norm” for a rule that is a norm. Using this terminology, it is not the case that local rules are of direct relevance. What is true, however, and what Raz in the block quote above would seem to agree with, is that local norms are of direct relevance.

VIII. THE RELEVANCE OF ADOPTED NORMS

By contrast, while adopted norms are of direct relevance within their own normative system, they are of mere indirect relevance within the system that adopts them.

In some cases, an adopted norm is relevant because it bears on the applicability of a local norm. In Saxby, there is a norm of English law that says that a debt is recoverable if it is incurred to play a lawful game. Bray J. had a duty to apply this local norm, if it was applicable. But was it applicable? To answer that question, Bray J. needed to determine whether Saxby loaned Brook money for a game lawful where it was played. To make that determination, Bray J. needed to apply Monte Carlo law. So, to work out whether the English norm applied to this case, Bray J. needed to first apply Monte Carlo law. Monte Carlo law is relevant to Saxby and Fulton's rights and duties, but only insofar as it mattered to the applicability of the debt-recovery norm.

In other cases, an adopted norm is relevant because it bears on the interpretation of a local norm. Recall Smith. The United States Supreme Court needed to interpret the term “to use” in the Criminal Code. The court thought the issue turned on the proper use of the term in ordinary contexts. Thus, to resolve the legal issue, the court relied on semantic norms. These norms were legally relevant—they helped in deciding Smith's liability to a criminal penalty—but indirectly, through their significance for the interpretation of a norm of direct relevance.

Similarly, in Copyright Owners, the High Court was tasked with interpreting a statute of the Imperial Parliament. The court reasoned that the Parliament would not intend to act contrary to a constitutional convention. There were two possible interpretations open to the court, one of which would include Australia within the statute's scope, one of which would not. On the first interpretation, the Parliament would have acted contrary to a constitutional convention. So, the court inferred that Parliament intended the latter interpretation. Thus, the convention was relevant to the rights and duties of the parties, but indirectly, via its relevance for the meaning of the statute.

In still other cases, an adopted norm is relevant to deciding on the exact consequences of a local norm. Consider a fresh example, this time of a game. During a break from music practice, Blake and Galloway and four friends began to “engage in some horseplay.”Footnote ⁴² Blake threw a bit of bark at Galloway's leg. Galloway picked it up and threw it back at Blake—striking him in the eye, causing serious injury. Blake sued Galloway, claiming that he had breached a duty of care to Blake. In Blake v. Galloway,Footnote ⁴³ Dyson L.J. said that the horseplay was “conducted in accordance with certain tacitly agreed understandings or conventions.”Footnote ⁴⁴ Those tacitly agreed norms included:

These norms of horseplay were not directly relevant: they did not in their own right establish any conclusion within English law. Rather, they were relevant in determining the exact content of the duty of care Galloway owed to Blake. In essence, Galloway had a duty not to act contrary to the horseplay norms. Because Dyson L.J. thought that Galloway had acted “in accordance with the tacit understandings or conventions of the game in which the claimant participated,”Footnote ⁴⁶ he held that Galloway was not liable for the injury he caused to Blake. Thus, the norms of the game were relevant to deciding on the parties’ duties and rights, but at one step remove from the norms of tort law.

Returning to Saxby, once Bray J. finds that the debt-recovery norm applies to the case, he is committed to reaching some conclusion on the basis of that norm. But what, exactly, does Saxby have a right to? How much money, exactly, does Fulton have a duty to pay him? Saxby loaned Brook £3,080 in 1905 and £1,070 in 1906. Saxby is entitled to the total of these amounts, plus interest. To work out the total, Bray J. needed to apply arithmetical standards of addition and multiplication.Footnote ⁴⁷ These arithmetical standards were indirectly relevant: they were relevant to deciding the parties’ rights and duties because they were relevant to deciding what, exactly, a norm of English law required.

Adopted norms can be indirectly relevant in more than one way. In Coughlan, the health authority's promise triggered the application of the doctrine of legitimate expectations. In addition, that promise figured in the determination of what, exactly, the doctrine of legitimate expectations required the health authority to do—in this case, to allow Coughlan to stay in her nursing home.

These are only six examples, but it is easy to see how to extend the analysis. Judges have regard to the norms of clubs and corporations and universities—when those norms shed light on the meaning or operation of local norms. Judges apply the promises and agreements in contracts—because they trigger the norms in the law of contracts. Adopted norms are legally relevant, but always and only indirectly.

IX. A DIFFERENCE IN RELEVANCE

Local and adopted norms can be distinguished by the directness of their relevance. Here, then, is my answer to the boundary question: the norms that are local to a legal system—i.e., part of a legal system—are directly relevant within it. By contrast, the norms adopted by a legal system are merely indirectly relevant within it.

Why should you endorse this test?

First, and most importantly, my test is extensionally adequate. It classifies all of the uncontroversial cases of adopted and local norms correctly, something that none of the existing tests do. Second, this test captures what I think is an intuitive thought about legal systems, namely, that the norms of that system have primary importance in determining rights and duties within that system. Third, the test is consistent with the most careful doctrinal thinking about law's boundaries.

Fourth, my test reveals a connection between adopted norms and something else of legal relevance that is not part of a legal system—namely, the facts of a case. The facts of a case bear on the meaning, applicability, or exact consequences of local norms. In these ways, they play auxiliary or facilitative roles in decision-making. Adopted norms play the same roles. It may be no coincidence, then, that conflicts-of-laws scholars say that matters of foreign law are matters of fact.Footnote ⁴⁸

Two further considerations in favor of the test I propose are that it shows the practical importance of the boundary question, and that it shows the respect in which each of Kramer's, Shapiro's, and Raz's tests are plausible. I explain both points in the next section.

X. PRACTICAL IMPLICATIONS

Dworkin says that “nothing important”Footnote ⁴⁹ turns on the boundary question. The question is merely “taxonomic,”Footnote ⁵⁰ he claims. If Dworkin means that it does not much matter what we use the word “law” to refer to, then I agree. But I take Dworkin to be making the more interesting claim that membership in a legal system is not of practical importance because relevance within a legal system does not track membership within it. Understood this way, Raz claims something similar. There are sometimes “procedural differences” relevant to “standards that are part of the law or merely enforceable according to law,” Raz says, but “much of the time the practical implications of a standard are the same either way.”Footnote ⁵¹ I think this claim is a mistake: it is often of great practical importance whether a norm is local or adopted. The test I propose can tell us why.

Compare two scenarios. The first is the real Saxby, in which a norm permitting roulette is adopted by English law. In the second scenario, this norm is also part of English law. There is probably no practical difference between these scenarios for Saxby and Fulton: either way, Saxby is entitled to his money back. But for anyone who wants to play roulette in England, there is a world of difference. If you play roulette in London, in the first scenario, you do not break the law; in the second, you do. It only makes sense to think that adoption does not matter if we narrow the discussion to the particular case in which a norm is adopted. Once we widen our perspective to take into account other cases, the practical importance of the question comes into focus.

Someone might object that I am not comparing like with like. With respect to Saxby, they will say: the content of the Monte Carlo norm is that, in Monte Carlo, roulette is permitted. Were a norm with the same content part of English law, it would not matter at all to would-be roulette players in London, because the norm would not apply to their activities. However, there is no reason to assume that the Monte Carlo norm is territorially bounded: many laws apply extraterritorially. Moreover, even when a law applies only within its system's territory, this limit will not be part of the content of that law. A general rule will establish the territorial extent of the norms of a system.Footnote ⁵² Whether such a rule is adopted in addition to the laws that it regulates is a contingent matter, one that is resolved differently under different conflicts-of-laws regimes.Footnote ⁵³

This example and others like it illustrate that the relevance of an adopted norm is closely circumscribed. If an adopted norm bears on the content or operation of a local norm, then the judge properly takes it into account. That is the case in the real Saxby. If an adopted norm does not bear on the meaning or operation of a local norm, then it has no legal effect. Because the norm of Monte Carlo law is not relevant to the application of English law in the London scenario, there is no cause to take it into account. By contrast, were it a local law, it would have legal effect, even if applying it was not a means of working out the content or significance of other local laws. So, whether a norm is adopted or local matters, practically speaking, because it makes a difference to the conditions under which it is relevant.

Once we see the practical importance of the boundary question, we can better appreciate what the existing answers to that question get right. If a norm is local, then it has a direct impact on people's rights and duties. Naturally, a political community will want to restrict such impacts to norms that it chooses. That is how the community ensures it is self-governing. Even more naturally, a community will want to deny such impacts to norms over which external actors have a decisive say. Otherwise it would be subject to an alien will. Shapiro and Kramer's proposals now take on a new significance. That a norm is not created locally, or can be manipulated externally, are both strong reasons not to give a norm direct effect in local law. By contrast, what Raz identifies are reasons—comity and personal autonomy—to give a norm indirect effect in local law. Although Kramer's, Shapiro's, and Raz's tests are presented as answers to the boundary question, they are more plausible as answers to a different, normative question, namely: Where should a legal system draw its boundaries? That is an interesting and important question—it is just not my question.

XI. OBJECTIONS

I imagine you might have several objections.

First, you might worry that it is circular to distinguish laws from other norms in terms of legal relevance. I agree that it would be circular—were we to need to know which norms are laws to know which norms are legally relevant. But we do not. We can know which norms help to make various propositions of law true without knowing which are laws. Indeed, the whole point of the boundary question is that the class of norms that are part of the truth conditions of propositions of law is larger than the class of norms that are laws.

The second objection is more challenging. My test gives us a way to distinguish local norms from adopted norms: only local norms are directly relevant. There is no difficulty distinguishing local rules from adopted norms: only adopted norms impose duties, confer powers, or grant permissions. However, in addition, there are surely adopted rules, i.e., rules that are locally relevant but that are not part of local law. For example, a Monte Carlo rule that defines “roulette” might be of relevance in working out Saxby's rights within English law. How, then, are we to distinguish local rules from adopted rules? Not by the directness of their relevance: both are merely indirectly relevant. Not by their character either: both are rules that are not norms. Given that my test is an answer to the boundary question, and the boundary question is about norms not rules, this worry does not present a frontal challenge. But it would reflect poorly on my test were it to provide no indication of how to distinguish local and adopted rules. Certainly it would leave us without a way to draw law's boundaries generally.

To show how to defeat the objection, it will help to have an example to work with. During the Cold War, the then-Ms. Karsov was arrested for anti-state activities in Poland. She was kept in terrible prison conditions, and her health steadily deteriorated. One of her friends, Mr. Szechter, offered to marry her, as a ruse to allow her to escape the country. Ms. Karzov agreed, the two married, and they fled to England. Once safely in London, the couple asked the High Court to declare the marriage null. In Szecther v. Szecther,Footnote ⁵⁴ the judge explained that under English law, “no marriage is valid if by the law of either party's domicile one party does not consent to marry the other.”Footnote ⁵⁵ Here, both parties were domiciled in Poland when they married. Under Polish law, consent is present only if the choice to marry is “unconstrained.”Footnote ⁵⁶ Because Ms. Karzov's choice was driven by fear, were the case to come before a Polish court, the marriage “would be held to be void.”Footnote ⁵⁷ Thus, under English law, the marriage was void as well. The judge accordingly issued a decree of nullity.

This case features two validity rules: one English, the other Polish. Neither is a norm. Both are indirectly relevant within English law: the Polish rule bears on the applicability of the English rule, which in turn bears on the operation of various norms of English law. What, then, sets the two rules apart? Here is one thing.Footnote ⁵⁸ The judge in Szecther v. Szechter treats the Polish rule as relevant because it is part of Polish law. What makes the rule part of Polish law is its relations with norms of Polish law (e.g., norms about the rights and duties of parties to a marriage). We know, on independent grounds (i.e., the test of directness of relevance), that the norms of Polish law are not part of English law. So, the Polish validity rule is relevant within English law in part because it is related to norms that are not part of English law. That is no part of why the English validity rule is relevant. Thus, the rules are not relevant for the same reason. In general, adopted rules are locally relevant (if they are locally relevant) in part because of their relevance for nonlocal norms; local rules are not relevant for that reason.

Finally, I imagine you might worry that my test is superficial. Although I have said what it is for a norm to be directly relevant, and I have said that local norms are directly relevant, I have not said how a norm comes to be directly relevant. I have not said by virtue of what facts or features a norm makes a direct difference to people's rights and duties. Without such an account, you might doubt that my answer to the boundary question provides a deep explanation of the boundaries of legal systems. Now, I am not sure that I must respond to this challenge. It seems to me that my test is no more superficial than any of the existing tests. Be that as it may, let me outline one possible response.

According to H.L.A. Hart's original theory of law, a norm becomes part of a legal system by virtue of being “recognised” under a rule of recognition.Footnote ⁵⁹ The fact of its recognition provides judges (among other law-applying officials) with a reason to apply or act on the norm—that is, to treat the norm as a basis for deciding on people's rights and duties. If a norm is picked out by a rule of recognition, then judges have a reason to treat the norm as relevant, independent of whether it bears on the meaning or operation of another norm. They have a reason to treat it as directly relevant, in other words. But this same reason establishes that the norm is part of the legal system. So, it establishes that the norm is of direct relevance. So, a norm is directly relevant because it is picked out by a rule of recognition.

If this is right, then it might seem to expose to me to another objection. Was this not a long way to read to find out that Hart was right, you might complain? Could I not have simply said that I agreed with Hart and skipped the last fifteen-odd pages? No, for three reasons. First, many philosophers have thought that Hart had no way to distinguish between (1) the norms of a legal system and (2) the norms that are not part of that legal system but that judges of that system have a reason to apply.Footnote ⁶⁰ The norms within (2) include the norms of foreign legal systems, games, clubs, and so on—all of the norms listed in Section I. Bray J., for instance, did as he ought to by applying norms of Monte Carlo law. This criticism of Hart is unsound, however. Judges have a reason to act on the norms of their own system because a rule of recognition picks those rules out. Judges have a reason to act on other norms, if they do, because doing so helps them to work out the meaning or implications of the rules of their system. The reasons in the two cases are different. So, Hart could have deflected the criticism. That might seem obvious at this point, but if so, it is because of the framework and examples set out earlier in this article. Second, I am not agreeing with Hart. I am simply invoking Hart's theory to show that we are not without resources to explain how a norm comes to be directly relevant. Finally, while my test may be consistent with Hart's theory, it does not depend on its correctness. Even if you do not agree with that theory—indeed, even if you are not a positivist—you can endorse my test.

XII. SUMMARY

We need to answer the boundary question to know what is a law of a system and what is not. But the existing answers to the boundary question are inadequate. Kramer believes that local but not adopted norms are free from nonlocal control. Shapiro says that local but not adopted norms are created locally. Raz, meanwhile, claims that judges rely on local and adopted norms for different reasons. On reflection, though, each proposal is vulnerable to obvious counterexamples. I have tried to take a different approach. I have looked beyond the philosophy of law to what doctrinal scholars say about law's boundaries. And I have tried to build up a test using a wide range of examples of judicial engagement with what everyone would accept are adopted norms. The resulting answer is that, while many norms are relevant to deciding on rights and duties within a legal system, only the norms that belong to that system are directly relevant. This answer has many virtues, including that it explains why leading legal philosophers favored other proposals. It also shows that the boundary questions is important, not only for our understanding of the nature of law, but also practically, for what it says about how rights and duties are determined.