THE HART–KELSEN DEBATE

Consider the following sequence of propositions:

Now, 2 clearly presupposes something like 3:

There is a logical sequence here: if there is a doubt about a statement of type 1, we would normally expect it to be resolved by an account of type 2.⁵

And if there is a doubt about 2, we would expect it to be resolved by an account of type 3. And then we need an explanation of what makes 3 true, and so we get to 4. This much, I take it, is common ground. But now a question that needs to be answered is this: Why is it the case that 4 has to be grounded in pointing to norms. Why could it not be something else?

Kelsen has a detailed answer to this question.⁶

The law, according to Kelsen, is first and foremost a system of norms. Norms are “ought” statements, prescribing certain modes of conduct. Unlike moral norms, however, Kelsen maintains that legal norms are created by acts of will. They are products of deliberate human action. For instance, some people gather in a hall, speak, raise their hands, count them, and promulgate a string of words. These are actions and events taking place at a specific time and space. To say that what we have described here is the enactment of a law is to interpret these actions and events by ascribing a normative significance to them. Kelsen, however, firmly believed in Hume's distinction between “is” and “ought” and in the impossibility of deriving “ought” conclusions from factual premises alone. Thus Kelsen believed that the law, which is comprised of norms or “ought” statements, cannot be reduced to those natural actions and events that give rise to it. The activity of gathering, speaking, and raising of hands, in itself, is not the law; legal norms are essentially “ought” statements and, as such, they cannot be deduced from factual premises alone.⁷

How is it possible, then, to ascribe an “ought” to those actions and events that purport to create legal norms? Kelsen's reply is enchantingly simple: we ascribe a legal ought to such norm-creating acts by ultimately presupposing it. Because “ought” cannot be derived from “is,” and legal norms are essentially “ought” statements, there must be some kind of an “ought” presupposition in the background rendering the normativity of law intelligible.

Thus an act can create law, Kelsen argues, if it is in accord with another, “higher” legal norm that authorizes its creation in that way. And the “higher” legal norm in turn is legally valid only if it has been created in accordance with yet another, even “higher” legal norm that authorizes its enactment. Ultimately, Kelsen argues, one must reach a point where the authorizing norm is no longer the product of an act of will but is simply presupposed, and this is what Kelsen calls the basic norm.⁸

According to Kelsen, then, it is necessarily the case that an explanation of type 4 must point to a master norm that makes it the case that certain acts of will create law and others do not. Without assuming such a norm, the normativity of the entire legal order remains unexplained. H. L. A. Hart seems to have concurred, with one crucial caveat: the master norm is not a presupposition, as Kelsen would have it, but a social norm, a social convention that people (mostly judges and other officials) actually follow. This is what the rule of recognition is: the social rule that a community follows, the rule that grounds the answer to the question of what makes statements of type 3 true or false in that particular society.⁹

But now, if you take Kelsen's question seriously, you should be puzzled by this. How can a social fact—that people actually follow a certain rule—be a relevant answer to Kelsen's question of what makes it the case that certain acts of will create the law and others do not? Crudely put, if you start with the question of how an “is” can generate an “ought,” you cannot expect an answer to it by pointing to another “is.” Has Hart failed to see this? Yes and no. Consider, for example, the game of chess. The rules of the game prescribe, for instance, that the bishop can be moved only diagonally. Thus when players move the bishop, they follow a rule. The rule undoubtedly prescribes an “ought”; it prescribes permissible and impermissible moves in the game. What is it, then, that determines this “ought” about rules of chess? Is it not simply the fact that this is how the game is played? The game is constituted by rules or conventions. Those rules are, in a clear sense, social rules that people follow in playing this particular game. The rules of chess have a dual function: they constitute what the game is, and they prescribe norms that players ought to follow.¹⁰

Similarly, Hart seems to claim, the rules of recognition define or constitute what law in a certain society is, and they prescribe (that is, authorize) modes of creating law in that society. Social rules can determine their “ought,” as it were, by being followed by a certain community, just as the rules of chess determine their “ought” within the game that is actually followed by the relevant community.

This cannot be so simple, however. In fact, the complications go both ways. Something seems to be missing from Hart's account, but something is missing from Kelsen's account as well. Let me begin with Hart. The obvious difficulty with the chess analogy is that the rules of the game are “ought” statements, in the sense of giving reasons for action, only for those who actually decide to play this particular game. To the extent that there is any normative aspect to the rules of chess, it is a conditional one: if you want to play chess, these are the rules that you ought to follow. But of course, you do not have to play at all, nor do you have to play this particular game. Leslie Green was one of those who observed this difficulty in Hart's account of the rule of recognition. As he put it, “Hart's view that the fundamental rules [of recognition] are ‘mere conventions’ continues to sit uneasily with any notion of obligation” and thus with the intuition that the rules of recognition point to the sources of law that “judges are legally bound to apply.”¹¹

Green is wrong to focus the problem, however, on the notion of legal obligations. Hart's account of legal obligations is sound as it is. The rule of recognition, just like the rules of chess, determines what the practice is. There is no particular difficulty in realizing that such rules have a dual function: they both determine what constitutes the practice and prescribe modes of conduct within it. The legal obligation to follow the rules of recognition is just like the chess player's obligation to move the bishop diagonally. Both are prescribed by the rules of the game. What such rules cannot prescribe, however, is an “ought” about playing the game to begin with. But that is true of the law as well. If there is an “ought” to play the game, so to speak, then this “ought” cannot be expected to come from the rules of recognition. The obligation to play by the rules—that is, to follow the law, if there is one—must come from moral and political considerations. The reasons for obeying the law cannot be derived from the norms that determine what the law is.¹²

But now one could wonder whether we have contradicted Kelsen at all. Have we not just conceded that the normativity of law, like that of any other conventional practice, has to be presupposed? When a couple of people sit down to play chess, they just presuppose that the rules of chess are those they are obliged to follow. In playing chess, they presuppose its normativity. And in “playing by the law,” lawyers, judges, and other participants presuppose the normativity of the legal order. This is basically what the concept of the basic norm is supposed to capture: the underlying presupposition of the normativity of the relevant practice. Is there anything more to it than a presupposition?

The problem is that even on Kelsen's own account, one can see that there must be more to it than a presupposition. Why is that? Because the specific content of any particular basic norm is crucially determined by actual practice. As Kelsen himself repeatedly argues, a successful revolution brings about a radical change in the content of the basic norm. Suppose, for example, that in a given legal system the basic norm is that the constitution enacted by Rex One is binding. At a certain point, a coup d'état takes place and a republican government is successfully installed. At this point, Kelsen admits, “one presupposes a new basic norm, no longer the basic norm delegating law making authority to the monarch, but a basic norm delegating authority to the revolutionary government.”¹³

Has Kelsen violated his own categorical injunction against deriving “ought” from “is” here?¹⁴

Possibly yes. The answer partly depends on the question of whether it is possible to separate the roles of the basic norm in answering the question of how we identify the law as such and in answering the question of law's normativity. An answer to the question of what counts as law or as law creating acts in a particular community cannot be detached from practice. As Kelsen himself basically realized, it is the actual practice of judges and other officials that ultimately determines what counts as law in their society. If the basic norm is a presupposition, it has to be an actual presupposition of particular people and, as such, a matter of fact. (Note that it does not matter for this argument exactly how we construe the idea of practice here—whether it consists of actions, propositional attitudes, or beliefs; all of these are within the “is” category.)

Kelsen may still insist, however, that even if the identification of law is practice-dependent, the concept of (legal) normativity still requires something like the presupposition of the basic norm. Perhaps there is something to it, but there are some problems here. To begin with, Kelsen makes it very difficult to understand what he means by the idea of a presupposition: on the one hand, he emphasizes that for any normative system to make sense as such, namely, as a normative system, one must presuppose its basic norm. On the other hand, he also emphasizes that this presupposition is a matter of choice: one can either adopt the basic norm or not. But then, how is this choice to be grounded? It cannot be grounded on the relevant normative system, since it is the function of the basic norm to presuppose its normative validity. Thus the impression one gets is that there are countless potential normative systems, such as law, religion, morality, and so on, that one can either accept or not just by presupposing their respective basic norms. But without any rational or objective grounding of such evaluative systems, the choice of any basic norm remains rather whimsical, devoid of any reason. It is difficult to understand how normativity can really be explained on the basis of such rationally groundless choices.¹⁵

Second, it is also doubtful that a distinction can be drawn between the question of legal validity and an account of law's normativity, as Kelsen seems to maintain. If a social practice determines what counts as law in a given community (as Kelsen admits), and laws are necessarily norms, then surely the practice determines how we identify laws as norms. There does not seem to be a separate question about how we identify or conceptualize the normativity of law. In short, even if we restrict the role of the basic norm to an explanation of the concept of systematic normativity, it is far from clear how helpful a tool it is. This is a complicated issue, however, and I will not try to substantiate this point here.

Whether there are rules of recognition is partly a matter of observation, not something that can be determined only by abstract argument. Nevertheless, there is an important lesson to be learned from the failure of Kelsen's antireductionism. The idea of the basic norm was intended by Kelsen to avoid a reduction of legal validity to social facts, precisely of the kind that Hart later suggested in the form of the rules of recognition. Kelsen thought that he could avoid such a reduction by insisting that the basic norm is a presupposition, not a social norm. But as we have seen, Kelsen's account of the basic norm actually violates his own antireductionist aspirations. Even if, in certain respects, the basic norm is a presupposition, its content is always determined by practice. The basic norms of, say, the U.S. legal system and that of the U.K. differ precisely because judges and other officials actually apply different criteria in determining what the laws in their respective legal systems are. The content of the basic norm is entirely practice-dependent. Once we see that this practice is rule-governed, namely, that in applying the criteria for determining what the law is in their legal systems, judges and other officials follow certain rules, it becomes very difficult to deny that there are rules of recognition more or less along the lines suggested by Hart.¹⁶

COORDINATION CONVENTIONS AND THE RULES OF RECOGNITION

Two questions remain: What kind of social norms are the rules of recognition? And to what extent do those rules shape our understanding of what the law is? A widely held view, reinforced, perhaps, by some of Hart's comments in his Postscript to The Concept of Law, maintains that the rules of recognition are social conventions more or less along the lines suggested by David Lewis.¹⁷

Lewis claims that conventions are social rules that emerge as practical solutions to wide-scale, recurrent coordination problems. A coordination problem arises when several agents have a particular structure of preferences with respect to their mutual modes of conduct; namely, that between several alternatives of conduct open to them in a given set of circumstances, each and every agent has a stronger preference for acting in concert with the other agents than his own preference for acting upon any one of the particular alternatives. Most coordination problems are easily solved by simple agreements between the agents to act upon one more or less arbitrarily chosen alternative, thus securing concerted action among them. However, when a particular coordination problem is recurrent, and agreement is difficult to obtain (mostly because of the large number of agents involved), a social rule is very likely to emerge, and this rule is a convention. Conventions, in other words, emerge as solutions to recurrent coordination problems, not as a result of an agreement, but as an alternative to such an agreement, precisely in those cases where agreements are difficult or impossible to obtain.

I have no doubt that many familiar types of social conventions are coordination conventions as Lewis explains them. However, as I argue elsewhere at some length, Lewis's analysis generalizes from some cases to all.¹⁸

There are, as we shall see below, other types of social conventions. But before we get to this, we need a more precise definition of what social conventions are and what makes them a unique type of social rules. There are two main features intuitively associated with conventional rules. First, conventional rules are, in a specific sense, arbitrary. If a rule is a convention, we should be able to point to an alternative rule that we could have followed instead, achieving basically the same purpose, as it were. Second, conventional rules normally lose their point if they are not actually followed. The reasons for following a rule that is conventional are closely tied to the fact that others follow it too. In fact, both of these intuitive features of conventional rules derive from a single, though complex, feature that I call here “conventionality,” defined as follows:

A rule R is conventional if and only if all the following conditions obtain:

1. There is a group of people, a community P, that normally follows R in circumstances C.¹⁹

   Note that following a rule (as opposed to just acting in accordance with a rule) is a complex condition. It probably entails that the agent regards the rule as a reason for action and perhaps a reason to exert pressure on others to comply with the rule, etc.

2. There is a main, or primary, reason (or a combination of reasons), call it A, for members of P to follow R in circumstances C.
3. There is at least one other potential rule S that if members of P had actually followed in circumstances C, then A would have been a sufficient reason for members of P to follow S instead of R in circumstances C.²⁰

   Assume that we rule out the possibility of backtracking conditionals and assume that the rules R and S are such that it is impossible to comply with both of them concomitantly in circumstances C.

A few brief clarifications are in place. The first condition indicates that conventions are social rules:²¹

A convention is a rule that is followed by a population. It is crucial to realize that not all rules have to meet this condition. The content of basic moral rules, for example, is not affected by practice of compliance. In other words, if N is the content of a rule, in morality it makes sense to say that N is entailed by “it ought to be that N.” However, “it ought to be that it is a convention that N” does not entail that “it is a convention that N.” Conventions must be practiced by a population in order to exist.

Second, it is not part of this condition of conventionality that members of P must be aware of the reason A in order to follow R. People may follow conventional rules for various misconceived reasons or, in fact, for no reason that is apparent to them at all. The conventionality of a rule does not depend on the subjective conception of the reasons for following the rule by those who follow it, nor does it depend on people's recognition that the rule is conventional.²²

The third condition explains the sense in which conventional rules are arbitrary. As Lewis himself emphasizes, it is crucial to note that arbitrariness (thus defined) should not be confused with indifference.²³

This condition does not entail that people who follow the convention ought to be indifferent as to the choice between R and S. The rule is arbitrary in the requisite sense, even if people do have a reason to prefer one over the other, but only as long as the reason to prefer one of the rules is not stronger than the reason to follow the rule that is actually followed by others.²⁴

Arbitrariness is an essential defining feature of conventional rules. A rule is arbitrary if it has a conceivable alternative. If a rule does not have an alternative that could have been followed instead without a significant loss in its function or purpose, then it is not a convention. Basic norms or principles of morality, for instance, are not conventions; properly defined and qualified, they do not admit of alternatives (in the sense defined above).²⁵

Admittedly, it is not easy to define what a conceivable alternative to a rule might be. Surely not every imaginable alternative to a rule would satisfy this condition. First, it has to be a rule that the same population could have followed in the same circumstances. Second, it has to be an alternative rule that is supported by the same reasons or functions that the original rule serves for the relevant population. Finally, in some loose sense that I cannot define here, the alternative rule has to be one that the relevant population can actually follow so that the cost of following it would not exceed the rule's benefits.

Lewis's account of social conventions easily meets these conditions. Rules that emerge as solutions to coordination problems are, indeed, arbitrary in this special sense, and it is pretty clear why the reasons people have for following coordination conventions are closely tied to the fact that others follow the same rule. But what about the rules of recognition: Are they conventions of this kind? There is some initial plausibility to an affirmative answer. After all, judges and other officials would need to follow those rules for identifying the sources of law that other judges and officials in their community also follow. Furthermore, as Coleman rightly notes, the coordinative rationale of the rules of recognition would also explain why judges and other officials have no particular incentive to defect; they would have no incentive to follow criteria of legality that differ from those followed by other officials.²⁶

There are, however, three main problems with the view that the rules of recognition are coordination conventions. First, this view misses the constitutive function of the rules of recognition—it misses the point that these conventions constitute to a considerable extent what law is. Second, the idea that the rules of recognition are coordination conventions is not easy to reconcile with the apparent political importance of these rules. Finally, the coordination conventions account blurs the distinction between the question of what law is and of what counts as law in a particular legal order. Let me explain these problems in some detail.

Consider, for example, the conventions that constitute the game of chess or those that constitute artistic genres such as theater or opera. To say that these are coordination conventions along the lines suggested by Lewis entails the assumption that first there was some recurrent coordination problem, and then a convention evolved as a solution to the problem. But this is implausible. Does it make sense to suggest that there was a coordination problem between potential chess players before chess was invented and now they play by the rules to solve the problem? Or that there was some coordination problem that the conventions of theater are there to solve, a problem that existed before theater evolved, as it were? I do not want to deny that there are some coordination problems that are solved by such conventions. The point is that the social function of solving a recurrent coordination problem is simply not the main rationale of many types of social conventions. That is so because prior to the emergence of those conventions, there simply was no coordination problem to be solved. Conventions evolve in response to a wide variety of social needs, and the need for coordination is just one of them. For example, conventions constitute various ways in which we express respect to one another, both linguistically and otherwise;²⁷

conventions constitute artistic genres, games, and numerous social institutions; social conventions often regulate interpersonal relationships in such settings as a workplace, or a party, or an academic gathering, and such. In short, conventions serve numerous functions by constituting ways in which people interact with each other and engage in socially valuable activities.

One might raise the following objection: even if we grant that not all conventions are necessarily coordinative, it is plausible to maintain that all social practices are. We could not have a social practice without some cooperative intention or attitude shared by the relevant participants. Bearing in mind two caveats that I mention below, I see no difficulty in this suggestion. It seems to me plausible to claim that social practices are sustainable only if the participants share some cooperative intentions or attitudes. The important point is, however, to keep these two issues separate; the main function of constitutive conventions is to constitute a social practice and regulate conduct within it. I have tried to argue that there is no need to assume that the main rationale of such conventions is to solve an antecedent recurrent coordination problem. This is quite compatible with the view suggested here that it is a condition of the existence of conventional (and other) practices that a certain cooperative aim is shared by its participants.

Now, one caveat that should be kept in mind is that social practices can exist and flourish even if their participants have different and incompatible conceptions of the point of the practices and their underlying rationale, at least to some extent.²⁸

Second, we should not assume that participation in a conventional practice is necessarily voluntary. Roughly, there are two kinds of conventional practices in this respect: those that potential participants need to opt into (like chess, for example), and those in which we are participants by default, as it were, and the only relevant question is whether we can opt out of them (or how costly opting out is). Examples of the latter would be the conventional practices that form our everyday social interactions, such as speaking a language or following basic rules of etiquette and civility. Consider, for example, the conventional aspect of language: some level of “opting out” is possible and in fact quite familiar from substandard uses of language. But typically such substandard uses are subject to criticism and hostile reaction, and there is a limit to how much one can deviate from the conventions. At some point opting out would amount to one's inability to use language. To conclude this point: even if we admit that some cooperative intentions or attitudes must be largely present in any conventional social practice for it to exist, we should be careful not to exaggerate this point and in particular we should not be tempted to assume that it entails voluntary participation.

As Hart himself seems to suggest, the rules of recognition are very much like the rules of chess: they constitute ways of creating law and recognizing it as such. Once again, it is not my purpose to deny that the rules of recognition solve various coordination problems. They do that as well. It would be a serious distortion, however, to miss their constitutive function. The rules that determine how law is created, modified, and recognized as law also partly constitute what the law in the relevant community is. They define the rules of the game, thus constituting what the game is.

There are two separate but closely related points here. First, we need to realize that social conventions tend to emerge as a response to many types of social needs, the need for coordination being only of them. Second, in many cases, the social conventions have a constitutive function in actually creating social practices that we are engaged in by following those conventions. Both of these observations apply to the rules of recognition. There is no reason to assume that the main function of such rules is to solve recurrent coordination problems, and it is important to bear in mind that the rules of recognition have a constitutive function in that they actually constitute the game, so to speak—they partly constitute what law is.

Why are these two points related? Basically, the reason is that coordination conventions tend not to have a constitutive function. If there is a recurrent coordination problem, and a social rule evolves to solve the problem for the relevant agents, in this the rule has basically exhausted its function. The reason for having the rule in the first place and the reason the agents have for following it in each and every instance are basically the same: to solve the relevant coordination problem. There is a trivial sense in which we can say that here, too, the rule constitutes a practice, namely, the practice of following that rule. But in this sense, every social rule that is actually followed is constitutive of the practice of following it.²⁹

This, however, is not the sense in which the rules of chess constitute a social practice. The game of chess does not consist in the practice of following its rules, though it is certainly constituted by it. People can sit down in front of a chessboard and follow the rules of chess without actually playing the game.³⁰ Chess is a very complex social practice; it is an elaborate social interaction that embodies certain conceptions of winning and losing, values related to what counts as a good game and a bad one or an elegant move and a sloppy one, and so forth. Following the rules is only part of this complex interaction. To be sure, I do not want to deny that the game is made possible by following its rules. On the contrary, as I have tried to suggest, the rules actually constitute the practice. But they do not exhaust it. The relation between the rules and their emergent social practice is not one of identity. There is no practice without the rules, and if the rules were different, the practice would be different as well, but there is more to the practice than just following its rules.³¹

At least one explanation for this non–identity relation consists in the complex social functions and needs that conventions tend to respond to. Chess has not evolved to solve a particular antecedent problem that we could identify irrespective of the game itself and of the more general human activity of playing competitive games. Chess can be understood only against the background of understanding a whole range of social needs and various aspects of human nature, such as our need to play games, to win, to be intellectually challenged, to be able to understand a distinction between real-life concerns and “artificial” or “detached” structures of interaction, and so forth. In other words, there are always some reasons (or needs, functions, etc.) in the background for having that kind of practice, and those reasons are closely entangled with the various values the practice instantiates. And then, once a conventional practice is in play, the practice may constitute further values that can be instantiated only by engaging in that practice. Once again, this complexity is typically not present in the case of coordination conventions. When the reason for having a social rule consists in solving an antecedent (recurrent) coordination problem, then following the rule to solve the problem is more or less all that there is to it.

If constitutive conventions tend to be responsive to a variety of human needs and values, as I claim that they do, you may wonder what makes such rules conventional at all. Basically, the answer consists in a combination of two facts. First, it is crucial to realize that the needs, functions, or values that such conventions respond to radically underdetermine the content of the rules that constitute the relevant social practice. Our need or desire to play some competitive intellectual board game does not determine the content of the rules of chess; the creative needs or desires to stage dramatic performances do not determine the particular theatrical genres that have emerged over time in various cultures; the reasons for showing respect to people under various circumstances do not determine the practices of etiquette that we have. Second, the reasons for following such constitutive rules are compliance-dependent reasons. The reasons for participating in a conventional practice crucially depend on the facts that the practice is there and its rules are actually followed by the relevant population. Without some level of general compliance (in the relevant population), there is no social practice. To the extent that anyone has a reason to participate, the reason must partly depend on the fact that the practice actually exists.

The combination of these two facts explains why constitutive rules of such practices are arbitrary and thus conventional. In other the words, the conditions of conventionality clearly apply here. Looking at the game of chess, for example, we can see that:

1. There is a population that follows a set of rules S₁ in circumstances C;
2. There are some reasons, call them A, for members of this population to follow S₁ in C; and
3. There are some alternative rules S₂ such that if members of that population had actually followed S₂ in circumstances C, then A would have been a sufficient reason for them to follow S₂ instead of S₁ in circumstances C.

Basically, the idea is that chess could be played differently, and if chess was played somewhat differently from the way it is, it would have been a sufficient reason for potential players to follow those rules that are actually being followed by others. That is why the constitutive rules of chess are conventional.

An obvious objection can be raised here: one can claim that if the rules of chess, for instance, were different from what they are, the game would not be chess but some other game. The identity of the game is substantially determined by the rules that constitute it; and if the rules are significantly different, the game is different as well.³²

I say “significantly” different because it is very doubtful that slight changes in the rules allow us to say that it would be a different game. Suppose, for example, that you have two communities of chess players: one plays the game with a rule that allows castling, while the other community plays it without this rule. Would we say that these two populations play a different game? I doubt it. We should say that they both play chess, albeit somewhat differently. However, if we imagine differences that are more and more substantial, at some point we would have to say that it is no longer the same game (though it may still be quite similar in its main functions and values)—which is just another way of saying that such practices, like chess, are constituted by rules.

Now, it is true that the identity of a conventional practice crucially depends on its constitutive rules. But this does not mean that the rules cannot be conventional. As long as we can envisage alternatives to the rules that are actually being followed, that is, without any significant loss in the purpose or function of those rules, and the reasons for following the rules are compliance-dependent, the rules are arbitrary and thus conventional in the requisite sense. In other words, we have to admit that there might be some difficult questions about what would count as alternative ways of doing roughly the same thing. But rough identity is all we need here. As long as we can envisage alternative rules that would basically achieve the same purpose for the relevant population or serve basically the same functions, and so on, we should have no problem characterizing the rules as conventional.

Realizing that constitutive conventions tend to emerge as responses to complex social and human needs and not just to coordination problems should make it much easier to understand why the specific conventions we happen to have may matter to us, sometimes a great deal. And the rules of recognition do matter, morally, politically, and otherwise. After all, it does matter to us who makes the law in our society and how it is done. The rules of recognition of legal systems are often politically important. Consider, for example, one of the most fundamental rules of recognition in the United States, namely, the rule that determines the supremacy of the U.S. Constitution. It should be easy to recognize that this is no trivial matter; it is something that most Americans feel strongly about, to say the least.³³

There are political and moral values associated with rules of recognition, values that it would be much less rational to attribute to rules that are simply there to solve a coordination problem. There are, of course, many coordination problems that it is very important to solve; but it is usually not very important how exactly we solve them as long as the solution is reasonably efficient.³⁴

Finally, the coordination account of the rules of recognition makes it unclear how these conventions of recognition relate to the concept of law. Consider chess again: without the conventions that constitute this game, there is no game of chess nor, consequently, a concept of chess. The rules of chess have a crucial role to play in constituting our concept of chess. On the other hand, if we think about a standard coordination convention, the picture is quite different. Consider, for example, a convention that determines on which side of the road to drive, or in which hand to hold the fork and the knife, or how to greet an acquaintance, or some such. In these cases we normally have the concept of the relevant activity irrespective of the conventions. In fact this is typically so, since the whole point of coordination conventions is to solve a problem that was there before the convention emerged, so it must be the case that we have a concept of the relevant activity irrespective of the conventions that have evolved to regulate it. Once again, it seems that law is more like chess than the coordination cases; without the social conventions that constitute ways of making law and recognizing it as such, it is difficult to imagine what kind of concept of law we could possibly have.

Perhaps this may not be the case? Hart comes close to suggesting that the concept of law is not really dependent on the rules of recognition. He does so when he considers (in a different context) the hypothetical of a “primitive” legal system that has no secondary rules and thus no rules of recognition.³⁵

Hart seems to suggest that such a legal order is possible, even though, as he argues, it would be defective and quite remote from law as we now understand it. So it seems that if we can imagine a legal system without rules of recognition, then the concept of law is not constituted in any significant way by such rules. Perhaps law is not like chess after all.

Let us reconstruct the exercise. Imagine “law” in a certain community that consists solely of primary rules, namely, a set of rules of conduct and no secondary rules whatsoever. Roughly, there are certain rules of conduct most people follow; they expect others to follow those rules and react in some hostile manner if they do not. So far, this set of social norms does not even come close to an example of a legal system. Almost any set of some rudimentary social norms fits the bill.³⁶

What makes it law? Nothing, as far as I can see. If there are no rules of recognition, how can these people tell the difference between following the law and any other set of norms they follow or think that they ought to follow. Suppose, to illustrate, that these people think they ought not to steal another's possessions. And then they also think people should greet a friend by a kiss on the cheek, never hold a knife in the left hand, pray to God twice a day, and show respect to their elderly. If this is a possible concept of law, it hardly resembles anything we would call “law” in our world. In other words, Hart's example of a “primitive” legal system is not legal in any meaningful sense; it is just a set of social norms. As Hart himself was at pains to argue, law is much more than this; and without the rules of recognition we could not possibly tell the difference.

Does this mean that the rules of recognition constitute the concept of law? In order to answer this question, we must distinguish between deep and surface conventions. I hope that once we clarify the distinction and the idea of deep conventions, the answer to this question will be easy to provide. This is the topic of the next section.

THE DEEP CONVENTIONS OF LAW

There are some reasons for having law—reasons that explain the main functions of law in our society—for example, the reasons to have some authoritative rules of conduct, the need to resolve conflicts in society, to create public goods, to solve collective action problems, and so on. And then there are, as we have seen, social conventions that determine what counts as law in a given community, namely, the rules of recognition. I want to argue that between the general reasons to have law and the local conventions that determine what counts as law in a particular legal system there is an intermediary layer of deep conventions, conventions that partly constitute the concept of law. The deep conventions of law are typically manifest in the surface conventions of recognition that are specific to any given society or legal system.³⁷

Deep conventions differ from surface conventions in the following ways:

1. Deep conventions emerge as normative responses to basic social and psychological needs. They serve relatively basic functions in our social world.
2. Deep conventions typically enable a set of surface conventions to emerge, and many types of surface conventions are made possible only as instantiations of deep conventions.
3. Under normal circumstances, deep conventions are actually practiced by following their corresponding surface conventions.
4. Compared with surface conventions, deep conventions are typically much more durable and less amenable to change.

Needless to say, this is all very schematic and needs to be shown. Let me follow the example of chess and explain what I mean here. The rules constituting chess are constitutive conventions. They constitute what chess is and how to play the game. In part, they constitute the point of playing the game and some of the specific values associated with it. But all this is possible only against the background of a deeper layer of shared understandings about what competitive games are. In order to play chess as a competitive game of a certain kind, players must follow a fairly elaborate set of norms, or deep conventions, determining the concept of games and the essential point(s) of engaging in such a practice. Chess, as a game of a particular kind, is only an instantiation of a more general human activity that we call “playing a (competitive) game.”

Admittedly, it is not easy to define a particular set of norms that constitute the activity we would call “playing a competitive game.” Nevertheless, some basic conventions are clear enough:

1. Playing a competitive game is basically a rule-governed activity. This means at least two main things: first, that in playing a game, the participants follow some rules; second, that the rules of the game define, among other things, what the game is and what counts as success or failure in the game, what counts as winning or losing, and so on. And of course these norms can be violated, so that putative players might break the rules or deviate from them in various ways.
2. Games involve a certain element of detachment from real-life concerns. The level of detachment varies considerably between different types of games and in different contexts and cultures. But even when the detachment from real-life concerns is minimal, games have a certain artificiality that is quite essential to our understanding of what games are. A violation of such norms typically involves a confusion; it typically manifests a misunderstanding of the situation.³⁸

   Consider the instances in which we say to someone “this is just a game”; it happens when people take the game too seriously, as it were; they get confused about this detachment norm.

3. Games have a fairly sharp demarcation of participants. Players are typically recognized as such and can be distinguished quite clearly from spectators, fans, and other nonparticipants. Again, these are norms that can be violated on particular occasions, for example, when fans or spectators attempt to participate as players and thus disrupt the game.

Let me clarify two points here. First, these three features are meant to be examples of deep conventions determining what games in our culture are. This list is not meant to be exhaustive, of course, and it is certainly not meant to be a definition of what (competitive) games are.³⁹

Second, I do not wish to claim that we must first have an abstract concept of games and then we can invent concrete instances of the abstract concept. This is not how our social and conceptual world develops. Abstract concepts emerge gradually, I presume, concomitantly with the particular cases that they instantiate. In any case, I do not purport to speculate about how people started to play games. In order to play a game like chess, I argue, participants (players and spectators) must also follow a set of background norms that determine what playing a game is.⁴⁰ Without such background norms of what, say, a competitive game is, the specific conventions constituting particular games like chess would not make sense; they would not be possible at all.

Now, the feature that makes it the case that these background rules are conventional is the fact they are basically arbitrary norms; the particular shape they take is underdetermined by the reasons that explain their emergence. And what makes them deep conventions is the fact that such norms are responsive to relatively deep aspects of human society and human nature and that they are normally practiced by following their corresponding surface conventions of the particular games we play.

Let me try to explain, beginning with the idea of depth. It is very easy to imagine a world in which people do not play chess, or soccer, or any particular game that is familiar to us. However, it is much more difficult to imagine a world in which people do not play any competitive games whatsoever. Or if we could imagine such a world, we would realize that it has to be very different from our own. Playing games is not something that we just happen to do; it is something that reflects some deep aspects of human culture and human nature.⁴¹

But now you may wonder: If deep conventions are responses to such deep aspects of our nature, what makes these norms conventional at all? The answer consists in the fact that the norms are underdetermined by the reasons, functions, needs, and so on that give rise to them. The same needs could be satisfied, as it were, by an alternative set of norms that would serve us just as well. Perhaps we could have some rituals that are not games but would serve similar functions; or we could have “games” that are not winnable; or we could have “games” that are not detached from real-life concerns, such as the Roman gladiators' “game” for their life.

Deep conventions tend to be rather elusive because they are typically practiced by following their corresponding surface conventions. We do not play competitive games in the abstract but particular games such as chess or soccer that are constituted by surface conventions. (It is worth noting that small children sometimes come close to playing an abstract game, as it were; they just play something and invent the rules as they go along. It is almost as if they practice what it is to play a game.)

Let me give another example from a different domain. In our culture there is an underlying convention that requires dress codes under certain circumstances, and then there are conventions about what counts as the appropriate outfit on particular types of occasions. The social practice appears to us in the practice of following the latter norms; what you see when you observe social behavior is the practice of following the surface conventions. But it is still the case that the deep convention is the underlying norm that people follow, albeit indirectly, that is, by following the corresponding surface conventions on the appropriate occasions. Suppose that the reasons or needs, functions, and so on for having dress code norms in our society are P. Let us assume that P consists in the reasons to show respect for people by some outward appearance. Now, it should not be difficult to imagine a different society where P is instantiated by a different kind of social practice, for instance, that people paint their faces in various colors in comparable circumstances (or perhaps they wear feathers or different sizes of earrings; the possibilities are numerous.) And then, of course, if you live in this different society, it would be pointless for you to associate the social functions of P with any particular dress code. This is what makes the underlying norms of dress codes conventional.

In other words, here is how the conventional setting works: in our culture, one way in which we manifest respect for people is by dressing in certain ways on certain occasions. In other societies, the same social function of showing respect on comparable occasions can be practiced by other means, such as painting one's face or wearing feathers, and so on. This is the relevant deep convention. But such deep conventions can be practiced only by following their corresponding surface conventions. So there might be a convention that if men attend a wedding they should wear a suit and a tie; this is how they are expected to manifest the relevant kind of respect for such an occasion, so this suit-and-tie norm is the surface convention men would follow on such occasions. And of course there are many other surface conventions that instantiate the same underlying deep convention (i.e., of showing respect by dressing in certain ways).

Deep conventions typically come to our attention when we face some deviant behavior or some doubts about the practice we are engaged in. My history professor at college used to say that when his king is threatened by checkmate, he would simply declare a republic. Taken seriously (which I am almost sure he did not), this is not just a violation of the rules of chess; it is a violation of the deep conventions constituting what counts as playing a competitive game.

Finally, note that surface conventions often come in layers with different degrees of shallowness, so to speak. Consider, for example, the deep conventions of representational art, say, in medieval Europe. As we know, they were instantiated by an elaborate set of surface conventions. But some of those surface conventions were probably deeper than others. I would guess that conventions of religious symbolism, composition, and perspective were deeper than specific conventions of, say, color symbolism (e.g., that blue represents virginity); and conventions of color symbolism may have been deeper than conventions about paint material or the size of the works, and so on. Similar degrees of shallowness are present in other cases. The deep conventions of theater, to take another example, are instantiated by surface conventions of particular genres of theater, and those in turn may be practiced by following even shallower conventions, say, about the number of acts, stage setting, and so on.

If this is the case, you may wonder whether there is a categorical distinction between deep and surface conventions at all; perhaps we are entitled to say only that there is a deeper than relation of conventions. In response, two points: first, it should be noted that deep conventions can rarely be followed on their own, as it were; deep conventions are actually practiced by following their corresponding surface conventions. When we play competitive games, we follow the deep conventions that constitute what competitive games are by following the surface conventions of particular games. Surface conventions, on the other hand, can be followed on their own even if there are further, even shallower conventions that people follow in those circumstances as well.

Another main difference between the deep and the surface conventions is that the deep is constitutive of the practice in ways that the deeper than is not necessarily. Without the deep conventions of theater (such as, for example, the convention about suspension of belief), there is no theater, at least not in any form that we are familiar with. Without the deep conventions of competitive games that we discuss above, there would not be a practice that we can call “competitive games.” The deep conventions constitute what the practice is. In contrast, the surface conventions that are deeper than others do not necessarily serve this constitutive function (though sometimes they may). For example, the conventions about color symbolism could easily be replaced with a different one (say, a different color or none at all) without any necessary relation to the deeper conventions about the religious significance of the work, composition, perspective, and so on. Surface conventions generally instantiate the deeper ones; they are different ways of doing that. Shallower conventions within the setting of other surface conventions do not necessarily instantiate the deeper ones. Their relation to the deeper conventions is typically more incidental.⁴²

Let us now return to law. The rules of recognition of the kind Hart had in mind are surface conventions. They determine what counts as law in a particular legal system or in a particular community. These surface conventions of recognition are instantiations of deep conventions about what law in our culture is. Roughly, the idea is that the rules of recognition relate to the deep conventions of law like particular games relate to deep conventions about what it is to play a competitive game. What would be the deep conventions of law?

Consider these three (hugely simplified) possible models of what the law is. According to one familiar conception, law is a product of the act of will of particular individuals or institutions. Let me call this the institutional model of law. At least two other models, however, are familiar from history: the customary model and the religious one. According to the customary model, law is not created by acts of will but by long-standing social customs; roughly, law is just those norms of conduct (or more likely some subset of them) that have been followed in the community for a long period of time. And then there is a third familiar model that is essentially religious: law is the expressed wish of God, grounded on the interpretation of some holy scripture such as the Bible or the Koran. To be sure, I am not claiming that these are the only models we are familiar with, and I am certainly not committed to historical accuracy here. I present these models here just to exemplify the fact that our deep conventions of law are indeed conventional, that they could be quite different from what they are.

Now, these three models exemplify very different conceptions of legal authority. They instantiate different conceptions of what the law is. As one should expect, they have a great deal in common. That is why they are different models of law; they form conventional solutions to similar problems and social needs, for example, the social needs to have mechanism for resolution of conflicts in society, to solve collective action problems, to produce public goods, and so forth. These and similar concerns constitute the basic reasons for having law and legal institutions in our societies. But these reasons, universal as they may be, can be instantiated by different sets of deep conventions. According to the institutional model, the one that more or less prevailed in the modern world, law is, by and large, the deliberate product of recognized and institutionalized authorities. According to religious models, law is the expressed will of God, not of human institutions. And then there were times and places where law was just a little bit of both of these, but mostly it consisted of the customs and traditions that have been followed for generations. The normative manifestations of these different models of law are examples of the deep conventions I have in mind. As with any other type of deep conventions, they are actually practiced by their corresponding surface conventions of recognition that are particular to the specific society in question. For example, a religious practice of law must have certain surface conventions that determine what counts as the relevant holy scripture, the “expressed” will of God, who gets to determine its interpretation, and so forth. Similarly, a customary model must have some surface conventions of recognition about what counts as a legal custom as opposed to norms about, for example, etiquette or desirable but not obligatory behavior, who gets to resolve interpretative questions about such matters, who gets to apply them to particular conflicts, and so on.

Let me sum up: the conventional foundation of law consists of two layers. There are deep conventions that determine what law is, and those deep conventions are instantiated by the surface conventions of recognition that are specific to particular legal systems. The concept of law is constituted by both layers of conventions. Our concept of law depends partly on the deep conventions that determine what we take law and legal institutions to consist in and partly on the specific institutions we have, those that are determined by the rules of recognition. Basically, this is just like chess. Without the rules of chess, we would not have a concept of chess. But we can have such a concept only because we already possess the deeper concept of playing competitive games, of which chess is just one instance. Both are profoundly conventional, and in this general insight I think that Hart was quite right.

Needless to say, this is all a very sketchy account. I am sure that more needs to be said about the precise nature of the deep conventions of law and about the ways in which the deep conventions are manifest in specific rules of recognition. However, there is one point I would like to develop a bit further here. There is, I think, an interesting difference between the ways in which deep and surface conventions are amenable to change. All social conventions change over time. Some of them dwindle and cease to exist; others get modified in substantial ways. But there is this important difference: surface conventions can be modified or abolished at will, and they often are, whereas deep conventions are much more durable and typically resist deliberate, institutional modification.

Surface conventions are often codified and thus replaced by institutional rules.⁴³

The rules of chess, for example, have been largely replaced by codified rules during the twentieth century.⁴⁴ Similar codification has occurred with many competitive sports. In other domains, such as the arts or the conventions of a natural language, codification is very rare, and surface conventions change much less institutionally. They typically change either by losing their point, so to speak, and then gradually fading away, or sometimes by a revolutionary change, one that brings about a new convention replacing the old one. All these modes of change are also present in the law. Rules of recognition sometimes fade away and cease to be followed; at other times they may be changed institutionally. The formation and development of the European Union, to mention a clear example, has brought about a considerable change in the rules of recognition of its member states. And then, of course, there are revolutionary changes as well.

Legal theorists often thought that it is a particular strength of Kelsen's account of the basic norm, compared with Hart's conventional rule of recognition, that it is capable of explaining the idea of a legal-political revolution. When a revolutionary change takes place in a certain regime, such as a coup d'état, Kelsen argued, the success of the revolution determines its legality.⁴⁵

A revolution is successful when judges and other officials actually presuppose a new basic norm and apply it instead of the basic norm of the old regime. If that change occurs, Kelsen argued, a new legal system is in place. As we have already noted, this observation is actually not a strength of Kelsen's account but reveals one of its main weaknesses. Realizing that it is the practice of judges and other officials that actually determines the content of the basic norm violates the antireductionist aspirations of Kelsen's theory. In addition, however, we can now see that there is nothing particularly mysterious or problematic in the idea of a revolution for a conventional account of law's foundations. Once we realize that the rules of recognition are surface conventions, it should be much easier to understand how they can be changed by a political revolution. Surface conventions, as we noted, often change through deliberate, institutional intervention. Changes that occur in the deep conventions, on the other hand, are much less obvious and typically much more profound.

Deep conventions typically change very gradually over a long period of time. The deep conventions we mentioned above, such as those that constitute our practices of playing competitive games or those that constitute forms of art such as theater and so on, have changed over time, perhaps even quite substantially, but those changes were very slow, spanning centuries. Sometimes, however, a radical change does occur even at the level of deep conventions. For example, the emergence of abstract art at the beginning of the twentieth century may have been such a radical change in the deep conventions of painting and sculpture. And it was a fairly rapid and quite a radical change. How such radical changes occur in deep conventions is a difficult question that I cannot really answer. My only purpose here is to point out that such changes are basically what we would normally call a social revolution. A political revolution, if successful, brings about a change in the surface conventions; in law these are the rules of recognition. Whether the revolution also amounts to a radical social change depends, I suggest, on the question of whether the deep conventions have changed as well.