NEGATIVE POSTIVISM AND THE HARD FACTS OF LIFE*

In his essay, "Negative and Positive Positivism," [Fn 1] Jules L. Coleman extends in two important ways the Legal Positivism of H. L. A. Hart. First, he shows that the "separability thesis" - the claim that no necessary or constitutive relationship exists between law and morality - to which Positivists are wedded does not entail the view, attributed by Ronald Dworkin to Legal Positivists, that law consists in "hard facts." Instead, the separability thesis requires only the possibility of deciding the truth of propositions of law. This weaker requirement can be met by rules of recognition that are semantically adequate but that fail to set out tests by means of which the legal status of rules and norms may be verified (pp. 143-48). Second, Coleman argues that one can subscribe to the separability thesis and still maintain that judges may have a legal duty to couch decisions in moral terms whenever the "content or proper formulation of [a] rule of recognition" is in dispute (p. 157). Such a legal duty may exist in virtue of a convergent practice "among judges of resolving the controversy in a particular way" (p. 159).

Coleman's claims are well-taken, and his discussions of the separability thesis and its entailments and of law as social convention put Legal Positivism on firmer ground. In this essay I will expand on Coleman's analysis in two ways. First, I will argue that in practice we are unlikely to find developed legal systems in which rules of recognition systematically fail to serve the epistemic function Coleman describes. That is, I will show that most often law is a matter of hard facts, and, given the many uses to which laws are commonly put, the situation could not practicably be otherwise. Second, I will contend that Coleman's analysis of convergent practices among judges forces us to redefine Hart's "secondary rules," i.e., the rules that regulate the interpretation and manipulation of primary rules of obligation in a legal system.[Fn 2] Once secondary rules are appropriately defined, the legal duty judges may have to apply moral principles when interpreting rules of recognition is subsumed within the class of legal duties that delimit the ways in which officials of all kinds may properly exercise discretion when working with secondary rules. In addition, the redefinition allows for the possibility that in a conceivable legal system every discretionary act taken by an official must be justified on moral grounds.

The nature and thrust of my claims will be more apparent if the main lines of Coleman's reformulation of Legal Positivism are clearly drawn. To this end, I will briefly summarize his position.

Coleman's modifications of Legal Positivism

At least since Hart's exchange with Lon Fuller in the Harvard LawReview,[Fn 3] and arguably since Bentham's Introduction to the Principles of Morals and Legislation,[Fn 4] a distinguishing feature of Legal Positivism has been its denial of a necessary or constitutive connection between law and morality. The most straightforward reading of this separability thesis is that offered by Coleman, according to whom it commits Legal Positivists "to the proposition that there exists at least one conceivable legal system in which the rule of recognition does not specify being a principle of morality among the truth conditions for any proposition of law" (p. 143). Coleman calls that form of Legal Positivism which only denies the existence of a necessary connection between law and morality "Negative Positivism" and claims that it is valid "just as long as the idea of a legal system in which moral truth is not a necessary condition of legal validity is not selfcontradictory" (p. 143). Coleman does not explain why one can without contradiction speak of a legal system in which truth as a moral principle is not a condition for the legality of any norm or rule, but it is easy to construct a case for his claim. It seems likely that Coleman agrees with Hart that legal systems are systems of social rules and that the conditions that are sufficient to establish the existence of social rules are inadequate to ground any claims about the morality of such rules. According to Hart, a rule is a social rule (1) if the behavior of those individuals subject to the rule indicates that the rule is usually followed, and (2) if the existence of the rule is used as a reason for doing what the rule demands, for criticizing deviant behavior, and for supporting demands that others comply as well.[Fn 5] These conditions may be met by an uncontroversially immoral social practice, say, by a rule a gang of muggers might adopt to restrict damaging and inefficient internal compeitition. Since only the presence of secondary rules distinguishes legal systems from other collections of social rules, and since a priori there are no moral restrictions on the form or content of secondary rules, one can imagine a legal system in which truth as a moral principle is not a condition for the legality of any rule or norm in the system.

Against this backdrop it is easy to follow Coleman's argument that Negative Positivism does not entail the view Dworkin attributes to Legal Positivists, namely, the view that law consists in hard facts. [Fn 6] Questions about laws are always questions about hard facts only if rules of recognition must be epistemically adequate, that is, only if rules of recognition must set out tests by means of which the legal status of other rules and norms may be
verified. For example, a rule of recognition which held that anything and only things enacted by Parliament are laws would be epistemically adequate: assuming that procedures for enacting laws are well specified, a rule or norm would be a law if Parliament had adopted it, and otherwise not. Dworkin holds Legal Positivists to the hard facts view in virtue of condition (1) for the existence of social rules, according to which the behavior of those persons who comprise the group in which a social rule is alleged to exist must indicate that the rule is usually followed. If a rule of recognition is not epistemically adequate, controversy over its implications may arise, in which case the required convergence of behavior among those who claim to follow the rule will be lacking. Hence, all social rules, and therefore all rules of recognition, must be epistemically strong (pp. 149-50).

If Dworkin was correct in asserting that Legal Positivists are committed to the hard facts view of law, he would succeed in undermining the claim that the social rule theory of law is descriptively accurate. It is obviously true that the implications of rules of recognition are occasionally controversial. But one can subscribe to the separability thesis and remain agnostic about the status of the epistemic sense of rules of recognition. Epistemic adequacy is not a condition for the existence of a social rule and is not entailed by condition (1). For example, it is both a social rule and a rule of law in the United States that federal officials must uphold the laws of that country. In the usual run of things this rule stimulates little controversy, but in a surprisingly large number of instances it is difficult to know just what behavior the rule requires. Uncertainty is especially likely to arise when different governing institutions advance diverse interpretations of federal laws, and, in principle, there may be no way to resolve all such disputes with finality. Even so, disagreement among officials over the entailments of the rule that the laws of the United States must be upheld does not establish that no such rule exists. To the contrary, disagreement over the propriety of acts presumes the existence of social rule of upholding the law. What is disputed is not the existence of a.rule, but how properly to go about conforming one's behavior to it. And since officials routinely adhere to the rule of upholding their country's laws and invoke the rule normatively in justification of their behavior and when criticizing the behavior of others, conditions (1) and (2) for the existence of a social rule are satisfied in this case even though the rule in question does not have a strong epistemic sense. Hence, contra Dworkin, Legal Positivists can maintain both that rules of recognition are social rules and that uncertainty may exist as to the implications of rules of recognition.

Legal Positivists are committed to the position that social rules, including rules of recognition, must be semantically adequate. A rule of recognition is semantically adequate if it sets out truth conditions for propositions of the form "X is a law in S," where S is the society that observes the rule of recognition, and X is a rule or norm which is alleged to be a law in the system the rule of recognition describes. For example, the rule "The law is whatever is morally correct" would be an epistemically inadequate but semantically adequate rule of recognition. Since morality is inherently controversial, the rule would not enable one to verify the legal status of other rules and norms. However, in light of this rule one could state that the proposition "X is a law in S" is true if the rule or norm X instantiates a true moral principle (pp. 140-43, 156-59).

Having shown that Legal Positivists are not driven to the hard facts view, Coleman considers whether judges may have a legal duty to apply moral principles in cases where the implications of rules of recognition are disputed, and, if so, whether the social rule theory of law can account for such a duty (pp. 156-62). He distinguishes two ways in which the implications of a rule of recognition may be controversial. First, as would likely be true of the rule "The law is whatever is morally correct," the content or meaning of a rule of recognition may be clear, but it may be unclear which norms or rules satisfy the condition the rule of recognition sets out. Second, the content or meaning of a rule of recognition may be in doubt, in which case "the question . . . pose[d] is not whether judges agree about which norms satisfy the same rule of recognition; rather, it is whether judges can be said to be applying the same rule" (p. 157). This second kind of uncertainty often arises when questions of law are not readily settled by referring to existing rules, but seem instead either to be proper subject matter for diverse, conflicting rules or to occupy the interstices between rules.

It is well known that Hart and Dworkin take opposing sides on the issue of discretion, with Hart arguing that, since discretion begins where rules end, there can be no such thing as a legal duty to conform discretionary acts to moral principles, and Dworkin claiming that, since judges feel bound to invoke principles of political morality when deciding hard cases, the account of rules of recognition as social rules must be mistaken. Coleman stakes out the middle ground, He contends that Hart is wrong to rule out the possibility that judges may have a legal duty to resort to moral arguments when deciding hard cases and that Dworkin is wrong to think that such a duty must exist in all conceivable legal systems.

Suppose a case arises under the rule "The law is whatever is morally correct." To resolve this dispute a judge would have to decide what in this case would be the morally proper thing to do. He might decide the case wrongly, but his legal duty to appeal to moral considerations when drawing a conclusion would be clear given the rule of recognition that governs the case. Since the rule "The law is whatever is morally correct" could be a rule
of recognition (nothing in the structure of social rules renders this impossible), Hart is wrong to claim that in principle judges cannot have a legal duty to invoke moral principles when the implications of a rule of recognition are in doubt.

Now suppose a case arises that calls into question the meaning of a rule of recognition, say, a case which requires a judge to decide which of two equally valid but conflictive laws holds sway. In Hart's language, this is a penumbral case that marks the limits of the convergent practice on which rules of recognition depend for their existence. That is, the case is a hard case in the sense that there does not exist a convergent practice among judges of according one law or the other superior status. Coleman agrees that penumbral cases may arise under any rule of recognition and that in such cases rules of recognition cannot account for any legal duties judges may have. Nonetheless, he allows for the possibility that in controversial cases of this sort judges may have a legal duty to take decisions that are informed by moral principles. Coleman can hold both positions simultaneously because he leaves open the possibility that, even though in a meaningful sense penumbral cases are outside the law, there may exist among judges "a practice of resolving ... controversies [over the meaning of rules of recognition] in a particular way," for example, a practice of appealing to principled moral arguments. If this is so, "then there is a legal duty even in controversial cases [of this kind]: A duty that does not derive from the principles judges cite," as Dworkin avers, "but from their acceptance of the practice of resolving these disputes by offering substantive moral argument" (pp. 159-60). Put simply, when penumbral cases call the meaning of rules of recognition into question, judges may have a legal duty to fix the meaning of those rules in a particular way, even though they may not have a duty to reach a particular answer. This will be so if within a population of judges there exists a social rule of resolving disputes over the meaning of rules of recognition by resorting to certain procedures, say, by appealing to moral considerations, by consulting poison-oracles, or by tossing dice. It is arguably true that judges in modern Western legal systems make a practice of giving moral reasons for discretionary judgments and that, therefore, judges suffer a legal duty to do so. This accounts for Dworkin's claim that judges have an obligation to give the right answer when deciding hard cases. However, since the existence of a convergent practice among judges of resorting to moral arguments when deciding penumbral cases is a contingent matter, Coleman is not committed to the strong thesis, to which Dworkin seems driven, that "judges must be under a legal obligation to resolve controversies arising in every conceivable rule of recognition by reliance on substantive moral argument" (p. 161). Unless it is construed as a truism, so that a system of social rules is not a legal system unless judges always have a duty to resort to moral arguments when fixing the meaning of rules of recognition, this strong thesis "seems straightforwardly false, since we can imagine immature legal systems (which are legal systems nonetheless) in which no practice for resolving disputes in the rule of recognition has yet developed - where all that succeeds is success. Or we could imagine the development of considerably less attractive practices for resolving such disputes, for example, the flip of a coin: heads, defendant wins; tails, plaintiff does" (p. 162). Thus, by sticking to Negative Positivism and refining our understanding of the degree to which social conventions may permeate legal systems, Coleman has strengthened the claim that the social rule theory of law is descriptively accurate and, in my opinion, has moved Legal Positivism closer to its final form.

Strategic interaction and the epistemic sense of rules of recognition

Coleman does not claim that we will ever find a legal system in which rules of recognition systematically fail to set out tests by means of which the legal status of rules and norms within the system may be verified. He shows only that it is logically proper, given the structure of social rules, to talk about a legal system in which no rule of recognition is epistemically adequate. I concur that it is possible to talk about such a system. My claim in this section will be that, given the uses to which laws are commonly put, we would have no use for a legal system in which even a substantial number of the rules of recognition were only semantically adequate.

Laws often facilitate strategic interactions.[Fn 7] I will use contracts and property rights to illustrate how laws peform this function. My example, exchange of economic goods, will instance a 2-person Prisoner's Dilemma (PD) and is therefore a social interaction of the interdependent sort. An interaction is interdependent if the size of one person's reward varies with the actions of another person. My discussion of this example and its related strategic representation will show that laws could not facilitate common strategic interactions in the absence of epistemically strong rules of recognition. The arguments I will present are not limited in applicability to solely economic exchange. Rather, they suggest how laws (and other social rules) facilitate all kinds of strategic interactions, including many that involve more than two persons.

Economic exchanges vary in quality along a continuum from discrete to relational. An exchange is discrete if "no relation exists between the parties apart from the simple exchange of goods."[Fn 8] In a perfectly discrete exchange, "I buy X from you at this minute at the most favorable price for which I can get it and we have no further relationship." [Fn 9] An exchange is relational to the degree that ties not present in discrete interactions connect the parties. Real economic exchanges contain many relational elements. For example, trading partners (as the name suggests) often do business with each other repeatedly and have an interest in maintaining an ongoing relationship. Warrantees, guarantees, return policies, and reputations also violate the conditions of discrete exchange because they extend aspects of present transactions into the future. And, obviously, the pooling of resources within a family or communal group is relational throughout.

Relational exchanges afford opportunities for strategic maneuvering not found in discrete exchanges. Each party to a discrete exchange has only a single, simple decision to make - to keep his good or to give it away - and each knows that his counterpart has only the same two options. If we assume that each prefers a good the other possesses to one he holds himself, an exchange would improve the well-being of both. If we also assume that both persons prefer larger amounts of the goods at stake to smaller amounts of those goods, then their exchange would have the incentive structure of a PD. Matrix I is a standard PD with labels appropriate to an economic exchange.

Matrix 1: Exchange as a Prisoner's Dilemma
 
 
  Player B
Yield 
Keep
 
Yield 
(3,3)*
 (1,4)
Player A
Keep
(4,1)
 (2,2)

*(Payoff to A, Payoff to B); Ordinal payoffs: 4 > 3 > 2 > I

As the matrix shows, trade would benefit both players (3 2). In spite of this, neither player has an incentive to exchange. Assume Player A decides to yield his good to Player B. B must decide whether to reciprocate. If she does, B will garner her second most preferred payoff (3), but refusing would afford her a bigger gain (4). Assume A decides to keep his good. If B yields in this case, she will receive the least desired payoff (1). If she also keeps her good, B assures herself at least (2). Thus, B has a dominant strategy of refusing to trade: she makes herself better off by keeping her good regardless of how she expects A to act. An identical logic confronts Player A. And if both A and B play their dominant strategies, they wind up in the lower right-hand corner of Matrix 1 (2,2). This outcome is an equilibrium: once there, neither player has an incentive unilaterally to change his choice of strategy. Thus, A and B are struck with an inferior outcome even though an exchange would make both of them better off.

Economically rational players, then, do not consummate discrete exchanges. Such players may, however, consummate relational exchanges, as the following example shows. Two individuals, a farmer and a fisherman, could profit from an exchange of goods. Moreover, the need of each for the other's produce is not limited to the present moment, which means that both could profit from a long-term trading relationship. Any exchange of goods that occurred between them would not, then, be perfectly discrete. This is so because both farmer and fisherman would know that the likelihood of profitable transactions in the future varies with the behavior of each in the present. For instance, each would risk losing the benefits of future trades by trying to make off with the other's produce without offering a satisfactory price or by trying to stick the other with shoddy goods. And, more importantly, given the opportunity for long-term gain, each might be willing to risk a small loss in the present as a means of getting an ongoing relationship started. [Fn 10]

In formal terms, repeated exchanges are iterated PD's, i.e., PD's that are played more than once. Given reasonable availability of information, a player in an interated PD can make his decision in future plays contingent on his counterpart's behavior in a present play. He can therefore use the threat of refusing to cooperate in the future (here, the threat of refusing to exchange) to protect himself against opportunism, and, as when a player employs a tit-for-tat strategy, he can, after demonstrating his resolve, be forgiving and attempt to persuade his counterpart that cooperation continues to be a feasible alternative to mutual defection. Thus, refusing to cooperate is not a dominant strategy in an interated PD. [Fn 11]

Not just any relational tie will encourage trade among economically rational persons, because not every relational aspect of exchanges strengthens the incentive to cooperate. Suppose, for example, I would like to buy from you an expensive object that requires a year to be manufactured. In the absence of effective legal contracts, we would not effect this exchange, and we might not do so even if we participate in an ongoing commercial relationship. You, as producer, would have no incentive to incur substantial reliance costs on solely the strength of my promise. In a year's time I might change my mind about wanting the good at the agreed price and renege on the bargain. That we had in the past engaged in several exchanges of inexpensive goods and had prospects for continuing that relationship might not allay your fears. Granted, I have proven myself trustworthy in small scale transactions. Does my reputation necessarily apply when the stakes are large? Similarly, I, as buyer, would be wary of compensating you in advance of production for fear that, once paid such a large sum, you would run off with the money. We might be able to overcome the problem of mutual distrust by breaking down the transaction into smaller parcels: I pay you part of the purchase price at regular intervals and, with equal frequency, you provide me with some of the good. However, if our exchange will not lend itself to this sort of structure, the deal will fall through.

In general, then, to stimulate trade relational features that strengthen the incentive to cooperate must be fostered and those that engender suspicion must be excised or accommodated. Legal systems often perform both functions. The law of contract, for example, makes possible the presentiation of future aspects of exchanges. [Fn 12] That is, contracts enable traders to handle in the present aspects of their transactions that extend into the future. And legally enforceable property rights reduce the risk of loss that exchange necessarily entails. After briefly explaining how rights in contract and property serve these ends, I will show that laws could not facilitate exchanges in the absence of epistemically strong rules of recognition.

Reconsider the examples discussed above. In one case exchange failed because no relational elements were present. As shown, economically rational players will not cooperate in once-played PD's. If we introduce laws that enforce property ownership, exchanges become relational because traders now hold legal rights and obligations against each other. In particular, they now know that opportunism, that is, theft, will meet with unpleasant consequences, and they know that others know this as well. In this way property rights change the structure of the payoff matrix and enhance the prospects for exchange. Matrix II shows this clearly.

Matrix 11: Exchange with Property Rights
 
Player B
Yield
Keep
Yield 
(3,3)
(1,4-P)
Player A
Keep
(4-P, 1)
(2,2)

P represents the legal penalty a player will suffer if he attempts to take advantage of his counterpart's good faith. (Nota bene: Ihave not assumed that the law compensates victims of theft. If victims receive damage awards, the payoffs to A and B in the upper right- and lower left-hand corners, repsectively, would have to change to reflect that fact.) At some level of P, the attractiveness of theft will fall below that of trade (3 > 4 - P). Even at this level yielding would not be a dominant strategy: a player could still wind up worse off by offering to give up his good (2 > 1). However, each player will know that if his counterpart decides to accept an offer, reciprocation is likely to ensure, for once a player is in receipt of another's good he can only make himself worse off by refusing to trade (4 - P < 3). Thus, property rights reduce the risk of making offers and increase the likelihood of exchange.

In the second case no transaction occurred because the relational elements present reduced the incentive to trade. You had no reason to begin production before being paid and I doubted the advisability of paying in advance. What is needed is a means to make our relational exchange discrete. If in the present each of us could be sure that the other will cooperate in the future, we would effect an exchange without delay. Contract is such a means. By entering into a legally binding contract, a party to an exchange commits himself to perform a specified act in the future, say, to pay an agreed purchase price or to deliver a purchased item. These commitments are believable (in part) because by reneging on an agreement a party will bring the weight of the law down upon himself. And, as shown above, if that weight is sufficiently heavy, reneging becomes an economically irrational course of action. [Fn 13]

Laws other than those that back property rights and contracts also serve strategic purposes. For example, laws that protect the physical security of persons make possible all sorts of social and economic interactions that otherwise would pose problems of collective action. Briefly, a self-interested individual would prefer that all members of his society except him be required to respect others' persons. This arrangement would provide both a high level of personal safety and the ability to exploit others for selfish gain. However, if a sizeable number of persons chooses in this self-interested way, the system of rights will fail and everyone will be worse off than he would have been had everyone, himself included, been forced to respect others' persons. [Fn 14] Traffic laws also facilitate strategic interactions. Every driver is better off in a regulated traffic system than in a system in which each driver individually must decide what to do. This is true even if the price of a regulated traffic system is an occasional wait at an empty intersection or an inability to park near one's place of business. Such short-term losses are outweighed by long-term gains to the collectivity and to individual drivers. Thus, traffic laws often facilitate strategic interactions by coordinating the independent decisions of large numbers of drivers. [Fn 15]

I must emphasize that my claim is only that laws facilitate strategic interactions. It is not true that all such interactions would fail in the absence of effective laws. Exchanges often occur simply because people participate in ongoing relationships of mutual advantage. In small communities where people interact frequently and rely on each other to support a division of labor, even without laws crimes may occur infrequently and exchanges that look risky ab extra may be consummated regularly. Traffic flows can organize conventionally without the assistance of authoritative institutions. For instance, people may drive on the right side of the road because they expect others to drive that way. [Fn 16] And, since people do not always act in economically rational ways, informal cooperation may emerge despite the incentive structure that characterizes an interaction. My claim is that even where strategic interactions would occur anyway, laws may reduce risks and augment the effectiveness of conventions, thereby smoothing out wrinkles; and where strategic interactions would otherwise fail, laws facilitate their occurrence for the reasons given above.

But laws would be less useful as devices for facilitating strategic interactions in the absence of epistemically adequate rules of recongition. The law of contract offers an obvious example. As shown, contracts facilitate exchanges because they endow parties with rights over the behavior of others. In contracting with you for the purchase of a good which requires a year to be manufactured, I gain the legal right to demand that you produce the good and I place myself under a legal obligation to surrender the agreed purchase price. [Fn 17] What must be made clear, however, is that our exchange succeeds not merely because we have rights and obligations against each other. Rather, the exchange occurs because each of us knows just what his rights and obligations are, knows how they came to exist and how they may be discharged, knows that courts will use established tests to verify their existence and will enforce them in conventional ways, and knows that, as a result, the other is likely to cooperate. Our exchange occurs because we know what the law is and how to use it to achieve desired ends.

Might contracts facilitate transactions even if rules of recognition were epistemically inadequate? The answer is "yes, but not as well." Suppose that the rule "The law is whatever is morally correct" regulates the legal system of which our contract is a part. Under this rule we know that, if a dispute arises between us, a judge will decide our case, not on the basis of the contract, but after considering whatever moral arguments we can put forward. This is not quite akin to saying that the case would be up for grabs because there may exist conventional ways of deciding disputes like that between you and me even though the rule of recognition that regulates our legal system is epistemically inadequate. If this is so, the legal resolution of our case may be predictable and the contract between us may serve its purpose. But there need not be such a convention and, even if one exists, our confidence that the terms of the contract will be enforced may be less than it would be if a suitable, epistemically strong rule of recognition was in effect.

Almost certainly, Hart had this argument, or something very much like it, in mind when he described rules of recognition as remedies "for the uncertainty of the regime of primary rules." [Fn 18] Absent rules of recognition, he wrote, "if doubts arise as to what the [primary] rules are or as to the precise scope of some [primary] rule, there will be no procedure for settling this doubt, either by reference to an authoritative text or to an official whose declarations on this point are authoritative." [Fn 19] Rules of recognition increase efficiency because they enable people to know with a high degree of certainty what their legal rights and obligations are. Hart continues, "Wherever . . . a rule of recognition is accepted, both private persons and officials are provided with authoritative criteria for identifying primary rules of obligation." [Fn 20] Only if we stress the epistemic sense of the word "criteria" will it be true that rules of recognition necessarily reduce inefficiency. As we have seen (and as Coleman readily admits), merely semantically adequate rules of recognition may not enable us to know what laws exist or what rights and obligations we possess in virtue of those laws.

Finally, although I cannot explore this topic in detail here, it seems true that in general rights and obligations that arise under social rules, for example, rights and obligations that stem from promises, will facilitate strategic interactions effectively only when the strong knowledge conditions described above are met. This suggests that wherever positive rights and obligations are widely employed in the course of social and economic interactions, the rules that undergird them will have a strong epistemic sense.

Self-referring secondary rules and the exercise of bounded discretion

Rules of recognition are a species of secondary rule. As mentioned, they set out conditions in light of which the status of other rules in the system may be assessed. Hart identifies two other kinds of secondary rule: "rules of adjudication" which establish procedures for authoritatively interpreting other rules; and "rules of change" which describe how new rules may be added to a system, old ones deleted, and those in need of modification amended. [Fn 21] Each type of secondary rule serves a distinct function and remedies a specific defect that would otherwise, that is, in the absence of rules of that kind, affect a legal system. But as secondary rules, the three types have one thing in common - they are all rules about "primary rules."

"Primary rules are concerned with the actions that individuals must or must not do." [Fn 12] Thus, primary rules establish the obligations and rights people have with respect to one another, e.g., the obligation not to harm and the right to personal security, and secondary rules describe procedures for the proper interpretation and manipulation of primary rules.

Recalling the discussion of the first section, judges may conventionally resolve disputes over the meaning of rules of recognition in particular ways, for example, by appealing to principled moral arguments or by tossing coins. These conventions are analytically distinct from the rules of recognition to which they apply and from secondary rules in general. Just as secondary rules are rules about other rules, these conventions are rules that govern the interpretation and manipulation of rules of recognition. That is, the convergent practices Coleman describes amount to secondary rules about one kind of secondary rule, namely, rules of recognition.

There seems to me to be nothing essentially mysterious about this. Indeed, a quick look at any developed legal system, or even at the constitution and by-laws of any reasonably sophisticated social organization, will establish that almost always officials must work within the confines of established procedures when interpreting and manipulating secondary rules of all kinds. For example, United States congressmen and senators are not free to alter by just any means the rules of parlimentary procedure that determine the amount of support a bill must have if it is to become a law. These rules can be changed, but only if certain procedures, detailed in other rules, are adhered to. Thus, it is uncontroversial that lawmakers can have legal duties to conform to established standards when interpreting and manipulating secondary rules. This, of course, immediately leads one to ask why, if legislators may have legal duties to manipulate secondary rules in particular ways, should we expect judges to be exempt from similar duties? The observation that legislators can enact rules to govern the handling of secondary rules while judges cannot provides too weak a foundation for such an expectation. Enactment need not be a condition for the existence of a secondary rule, and, obviously, it could not be a condition for the existence of every secondary rule. It is only because students of jurisprudence, Legal Positivistis and others alike, preoccupy themselves with analyses of judicial behavior that obvious similarities between judges and other lawmakers are ignored, as are similarities among the ways that different types of secondary rule may be modified and interpreted. Once attention is paid to these similarities, the observation that judges may have a legal duty to interpret rules of recognition in particular ways, for example, by appealing to principled moral arguments, is rendered pedestrian.

As anyone who has read The Concept of Law will know, Hart talks at length of rules that restrict the actions even supreme lawmaking bodies may legally take. [Fn  23] Often such rules have a self-referential aspect. For example, it might be a constitutional provision that a two-thirds majority of both houses of a bicameral legislature is required to amend any portion of the constitution, including the provision that sets out the two-thirds majority rule. It is therefore surprising to discover that Hart describes secondary rules in a way that completely ignores this obvious and important characteristic of such rules. Hart writes that secondary rules "specify the ways in which . . . primary rules may be conclusively ascertained, introduced, eliminated, varied, and the fact of their violation conclusively determined." [Fn 24] This description fails to allow for the possibility that secondary rules may establish procedures for working with secondary rules, including procedures for working with those secondary rules that describe the procedures. In other words, in Hart's system there are no rules about secondary rules. This seems to me an oversight and an important one because in some cases the fact that a secondary rule has a self-referential aspect is the only thing that explains the legal duty lawmakers have to conform their behavior to an established standard when manipulating a secondary rule. This would be the case, for example, if legislators desired to amend the constitutional provision described above so that only an ordinary majority of both houses was required for the passage of future constitutional amendments. Given the self-referential aspect of the original provision, a two-thirds majority would be required to amend the consitution in this way, and such an extraordinary majority would be required only because the original provision described a procedure by means of which it, along with other provisions, could be changed.

This oversight is all the more surprising because Hart endorses the possibility of self-referring laws and, even though he does not explicitly give them this extension, his arguments for the possibility of self-referring laws support the possibility of self-referring secondary rules.[Fn 25] Fortunately, a minor change in Hart's description of the function of secondary rules will allow for the possibility that secondary rules may refer to themselves: Secondary rules specify the ways in which primary rules and secondary rules may be conclusively ascertained, introduced, eliminated, varied, and the fact of their violation conclusively determined. The added phrase broadens the compass of secondary rules so that they may now refer to primary rules, to other secondary rules, and to themselves. The convergent practices among judges that Coleman describes are, then, secondary rules about secondary rules, and they account for the duties judges may confront whenever they manipulate secondary rules, including the secondary rules that are the source of those duties.

The self-referential quality of secondary rules is important for another reason. It entails the possibility that in a conceivable legal system duties may restrict the ways in which any lawmaker may permissibly employ his discretion. That is, discretion free of legal restraint may not exist in some actual or possible system of laws. It need not be true that in this system lawmakers would always or ever be bound to apply moral principles when fixing the meaning of secondary rules or attempting to straighten out their implications; they might be required always to roll dice instead. Still, lawmakers would have to conform every use of discretion to some established standard. Thus, the modified definition of secondary rules frees Legal Positivism from any commitment to the existence of legally unbounded discretion in every conceivable legal system, even though it leaves open the possibility that in any actual system of law discretion and duty may not go hand in hand.

Conclusion

Consistent with what I take to be a distinguishing feature of Legal Positivism, I have maintained throughout this essay that it is possible to distinguish the structure of social rules and, hence, the structure of legal systems, from the uses to which social rules and laws are put. Thus, I have argued, along with Coleman, that Legal Positivists are not committed to the view that law consists in hard facts because the structure of social rules does not entail that such rules must have a strong epistemic sense. Likewise, the observation that judges and lawmakers in general may have legal duties to conform their behavior to established standards when the meaning of secondary rules is in doubt is consistent with, though not entailed by, the social rule theory of law, and Legal Positivists can admit as much without committing themselves to the stronger claim that such duties must exist in every conceivable legal system.

But, in the struggle to straighten out the entailments of the social rule theory of law, neither Legal Positivists nor anyone else should lose sight of the practical validity of Hart's claim that secondary rules remedy defects in regimes of primary rules and that they accomplish this feat because they have a strong epistemic sense. We often put secondary rules to uses they could not serve if they were only semantically adequate. And since most often we are more concerned with the uses to which laws are put than with the structure of legal systems, there is something to be said for worrying more about what the law actually looks like than about what it could possibly be.

Charles Silver

NOTES

* I would like to thank Russell Hardin and Bart Schultz for helpful discussions of the issues I take up in this essay. I would also like to thank Jon Elster for his help, though I confess to having been unable to come to grips with an indictment he raised of the claim that purely semantically adequate rules of recognition could always be social rules. Finally, Jules Coleman provided valuable and frank advice concerning both the exegesis of his position and the strengths and weaknesses of my arguments.

  1. 11 Journal of Legal Studies (January 1982), pp. 139-64. Page references are to this article.

  2. The Concept of Law (Oxford; Clarendon Press, 1961), pp. 90ff (hereafter cited as COL).

  3. H. L. A. Hart, "Positivism and the Separation of Law and Morals," 71 Harvard Law Review (1958), pp. 593-629.

  4. New York: Methuen & Co., 1982.

  5. COL, pp. 77-89. Hart refers to condition (2) as the requirement that social rules have an "internal aspect" in addition to the regularity of behavior that characterizes social habits. "If a social rule is to exist some at least must look upon the behaviour in question as a general standard to be followed by the group as a whole. A social rule has an 'internal' aspect, in addition to the external aspect [what I have called condition (1)] which it shares with a social habit and which consists in the regular uniform behavior which an observer could record" (COL, p. 55).

  6. Here and throughout I shall follow Coleman's description of Dworkin's position ("Negative and Positive Positivism," pp. 143ff). I do so because Coleman gives Dworkin's arguments as plausible a reading as I have been able to find, indeed, a more plausible reading than I think Dworkin's arguments usually merit. But see Dworkin's reply to Coleman's article in Ronald Dworkin and Contemporary Jurisprudence, ed. Marshall Cohen (Totowa, N.J.: Rowman & Allanheld, 1983) for Dworkin's thoughts on the accuracy of Coleman's constructions.

  7. See Lon L. Fuller, "Law and Human Interaction," in The Principles of Social Order, ed. K. 1. Winston, (Durham, NC: Duke University Press, 198 1), pp. 211-246, and other papers in this volume. In general, Fuller had a clear understanding of the uses to which laws are commonly put in social interactions and of the restrictions this imposes on the enterprise of adjudication.

  8. Ian R. Macneil, The New Social Contract: An Inquiry into Modern Contractual Relations (New Haven, CT: Yale University Press, 1980), p. 10.

  9. Russell Hardin, "Exchange Theory on Strategic Bases," 21 Social Science Information (1982), 255. See also Victor Goldberg, "Toward and Expanded Economic Theory of Contract," 10 Journal of Economic Issues (1976), pp. 45-61.

  10. The rationality of refusing to cooperate in a one-shot PD is well established. See Russell Hardin, Collective Action (Baltimore: Resources for the Future, 1982), for a discussion and bibliographical information.

  11. The most recent discussion of tit-for-tat is Robert Axelrod, The Evolution of Cooperation (New York: Basic Books, 1984), ch. 1.

  12. The return to currency of the word "presentiate" is due to Macneil, The New Social Contract, and other essays by Macneil on the law of contract.

  13. Damage rules alter the rationality of trade in ways additional to that described. Indeed, their chief effect may be to render contractors indifferent between the performance of contracts and the breach of contracts accompanied by the payment of damages. I will examine this topic in more detail in a subsequent essay on the morality of contract.

  14. Russell Hardin sets out the analytics of the problem of collective action in "Collective Action as an Agreeable n-Prisoner's Dilemma," 16 Behavioral Science (September 1971), pp. 472-81. And, of course, there is Mancur Olson's classic, The Logic of Collective Action (Cambridge, MA: Harvard University Press, 1965). I provide a strategic representation of the problem in "Utilitarian Participation," 4/5 Social Science Information (1984).

  15. Thomas C. Schelling discusses lots of examples of this kind in his various articles, many of which have been collected in Choices and Consequences (Cambridge, MA: Harvard University Press, 1984), and in his books, The Strategy of Conflict (Cambridge, MA: Harvard University Press, 1960) and Micromotives and Macrobehavior (New York: W. W. Norton, 1978).

  16. See David K. Lewis, Convention: A Philosophical Study (Cambridge: MA: Harvard University Press, 1969) for a discussion of the informal rise of social coordinations.

  17. Many legal systems no longer require specific performance. See P. S. Atayah, An Introduction to the Law of Contract 3rd ed. (Oxford: Clarendon Press, 1981), p.139. However, I do not think that this simplification causes any problems.

  18. COL, p. 92.

  19. COL, p. 90.

  20. COL, p. 97.

  21. COL, pp. 89-96.

  22. COL, p. 92.

  23. COL, pp. 64-76.

  24. COL, p. 92.

  25. "Self-referring Laws," in H. L. A. Hart, Essays in Jurisprudence and Philosophy (Oxford: Clarendon Press, 1983), pp. 170-78. This essay came to my attention only after I had submitted this article for publication.