GOVERT DEN HARTOGH A CONVENTIONALIST THEORY OF OBLIGATION ? My aim in this paper is to contribute to the elaboration of a conventionalist theory of law. My starting point is the institutional theory, as it has been developed primarily by Herbert Hart. According to Hart, law is a system of primary and secondary rules; the secondary rules govern the identification and functioning of legislative and judiciary authorities. On two questions the answers provided by the theory are less than satisfactory. The first question is what it means for a social rule to exist, the second what it means for such a rule to have an obligating force. Conventionalism purports to offer a remedy of the first defect; it undertakes to explain how rules can emerge and stabilize in contexts of interdependent action. I believe that it has not yet been completely successful in this enterprise. The conventionalist theories which have been developed so far also have not been able to explain what it is for at least some of these rules to have an obligating force. The reason for both shortcomings is, as I hope to show, that any attempt to give the required explanations seems to involve itself in a vicious circularity. This is the problem I will try to solve. 1. PROMISING AND EXPECTATIONS The problem of circularity I have in mind is a rather common one in social theory. It can perhaps be pointed out most clearly in the socalled indirect theory of the obligation of promises. What a promise, according to this account, is intended to produce directly is not an obligation, but something which, under appropriate circumstances, brings an obligation along. Indirect explanations diverge somewhat ? This is a substantially revised version of a paper published before in Nederlands Tijdschrift voor Rechtsfilosofie en Rechtstheorie 24 (1995), W. E. J. Tjeenk Willink, pp. 8–25. Law and Philosophy 17: 351–376, 1998. © 1998 Kluwer Academic Publishers. Printed in the Netherlands. 352 GOVERT DEN HARTOGH in their identification of this “something”. I will proceed on the assumption that if I promise you to do x, what I intend to induce in you is only the expectation that I will do x.1 An alternative would be to suggest that I also want you to act on this expectation. But the expectation is common to both alternatives, and this is the element that causes the problem. In continental contract law, the dominant view is the direct theory which explains the obligation as a direct result of an act of the will which is only declared by the promise.2 “Explaining”, however, is something of an euphemism. For it remains rather mysterious how it is possible to make an obligation exist by simply willing it to exist.3 1 The promisor directly states his intention to do x or makes an assertion about his doing x in the future. Páll S. Ardal, “And That’s a Promise,” Philosophical Quarterly 18 (1968), pp. 225–237; G. J. Warnock, The Object of Morality (Methuen: London, 1971), pp. 101–111; Jan Narveson, “Promising, Expecting, and Utility,” Canadian Journal of Philosophy 1 (1971), pp. 207–233; Oswald Hanfling, “Promises, Games and Institutions,” Proceedings Aristotelian Society 75 (1974/5), pp. 13–31; R. Fogelin, “Richard Price on Promising: a Limited Defense,” Journal of the History of Philosophy 21 (1983), p. 289. Or he communicates his intention to induce the promisee to rely on his doing x. Neil McCormick, “Voluntary Obligations and Normative Powers,” Aristotelian Society Suppl. Vol. 46 (1972), pp. 59–78; and Legal Right and Social Democracy (Oxford: Oxford University Press, 1982), pp. 190–211; Albert Weale, “Consent,” Political Studies 26 (1978), pp. 65–77; P. S. Atiyah, Promises, Morals, and Law (Oxford: Oxford University Press, 1981); Christopher McMahon, “Promising and Coordination,” American Philosophical Quarterly 26 (1989), pp. 239–247. Indirect theories are not necessarily utilitarian, cf. Thomas Scanlon, “Promises and Practices,” Philosophy and Public Affairs 20 (1991), pp. 199–226. 2 For Dutch contract law see, e.g., A. S. Hartkamp, Verbintenissenrecht II, Algemene Leer der overeenkomsten volgens het Nieuw Burgerlijk Wetboek (Zwolle: Tjeenk Willink, 1985). On inspection his “theory” turns out to consist of nothing but the bare assertion that the (declaration of) the will of the parties is the criterion (ratio essendi et cognoscendi) of the existence and extent of their obligations. It is only confirmed by reference to legislative and judicial authority. 3 Attempts to explain this mystery have been made by Joseph Raz, “Voluntary Obligations and Normative Powers” (Reply to McCormick), Aristotelian Society Suppl. Vol. 46 (1972), pp. 80–103; and Michael H. Robins, Promising, Intending, and Moral Autonomy (Cambridge: Cambridge University Press, 1984). The most promising way to develop a volitional theory probably would be by exploiting the notion of a “strategic intention”, i.e., an intention to do something, not because you have reason to do it, but because you have reason to intend to do it. David Gauthier, Morals by Agreement (Oxford: Oxford University Press, 1986), ch. 6; A CONVENTIONALIST THEORY OF OBLIGATION 353 If only for that reason, it is worthwhile to consider the merits of the indirect theory. Another reason is that it may be more able than the volitional theory to make sense of contract law as it actually exists. Promising, on any account, is a form of communication analyzable in a Gricean manner. According to the indirect theory I intend you to expect me to act in a certain way in accordance with your interest, and I intend you to recognize this first intention and to take it as (part of) your reason to start expecting. But when you trust me to act in a certain way, and especially when you permit your actions to be informed by this trust, letting you down would be a breach of trust and hence morally wrong. This is the way expectation leads to obligation. The obligation depends on a principle of keeping faith. The interest which is protected by the obligation is your interest in being able to adjust your plans to my future actions. I believe that these aspects of the indirect explanation make it intuitively appealing; that is another reason why its seeming circularity is troubling. Why should you believe, and I expect you to believe, that I will act in a way conforming to your interest? Perhaps, having made the promise, it will also be in my interest to act in that way. In that case the ascription of an obligation to me does not add significantly to the justification of your expectations of my behavior. Leaving out of account every aspect of a moral nature we could solve our coordination-problem quite happily. And this seems indeed to be the correct description of a major part – the “low promising”4 part – of the practice of promising. We agree to meet each other, and each of us acts in his own interest by meeting the agreement first. We therefore expect each other to meet the agreement, but we do not hold each other to lie under an obligation to do so. But, to take the second horn of a dilemma, it could also be that acting in a way conforming to your interest would not conform to mine. Why, in that case, would you expect, and I expect you to expect, me to act in that way? The obvious answer is: because Edward F. McClennen, Rationality and Dynamic Choice (Cambridge: Cambridge University Press, 1990), chs. 9 & 12; Michael Robins, “Is it Rational to Carry Out Strategic Intentions?” Philosophia 25 (1996–7). See infra, at note 42. 4 F. S. McNeilly, “Promises De-Moralized,” Philosophical Review 81 (1972), pp. 63–81; cf. David Lewis, Convention: A Philosophical Study (Cambridge, Mass.: Harvard University Press, 1969), pp. 83–88, 188–189. 354 GOVERT DEN HARTOGH you believe me to be under an obligation to do so. That is, indeed, the characteristic manner obligations function in social interaction: bridging the gap between partially divergent interests and thereby enabling agents to avoid sub-optimal outcomes. This answer, however, seems to undermine our suggested explanation. For the obligation was first presented as a product of your expectations; but now it seems that your expectations can only be justified by the ascription of a prior obligation. That is the circularity.5 Why should I bother to induce in you an expectation that I will do something which I will really have no reason to do at all? Surely because if you act on your expectation, I will benefit. What I am trying to do is to assure you that in that case I will take no advantage. But how can I succeed in this attempt without committing myself in any way? A similar problem of circularity occurs in the English legal doctrine of consideration. A unilateral promise which is not in payment of a benefit enjoyed, and on which the other party did not act to his detriment, is not enforceable in English law. But mutual gratuitous promises are. When such promises are exchanged, both conditional on the future performance of the other party, each promise is held to be valid because of the other. Counter-promise is judged to be a good “consideration”, though not quite as good as benefit or reliance.6 However, either the counter-promise is already valid or it is invalid. In the first case the consideration is redundant, in the second case unjustified. We might attempt to solve these problems by extending the lowpromising account in the following way. Most contracts are made with parties we may expect to meet again in the future in similar circumstances. In that case the parties will do well to follow a condi5 Noticed by H. A. Prichard, “The Obligation to Keep a Promise,” in Moral Obligation (Oxford: Oxford University Press, 1968), pp. 169–179; G. E. M. Anscombe, “Rules, Rights and Promises” (1978) in Ethics, Religion and Politics, Coll. Philos. Papers, Vol. III (Oxford: Blackwell, 1981), pp. 97–103; Robins (1984), supra, note 3, at 9, 133; Joseph Raz, The Morality of Freedom (Oxford: Oxford University Press, 1986), pp. 85, 95–96. Atiyah, supra, note 1, at 64–69, argues that even if the promisee could be shown to be justified in his expectation, it would not have been explained why the harm caused by his detrimental reliance creates a liability in the promisor. 6 Atiyah, supra, note 1, at 3, 37–38. A CONVENTIONALIST THEORY OF OBLIGATION 355 tional strategy: not to join in with a person who did not fulfil his part of the bargain on the last occasion (or the last n occasions). Then delivering the goods and relying on the fact actually is a stable equilibrium outcome; it is in the interest of each party to do his part. Everybody acts “on the security of never being trusted again”,7 in order to keep up his reputation. We can be led by the same motive when dealing with a party who we know we will probably never meet again, but who belongs with us to the same small group of frequently interacting partners (merchant bankers, diamond dealers, national states).8 What kind of a reputation is at stake here? It cannot properly be the reputation of a person to be trusted. For to trust a person in a full, non-etiolated sense means to rely, not on his ability to consider his own long-term interests, but on his being prepared to reckon with the interests of others, or to honor their trust. Relying on the desire to stand in good repute does not introduce any element of a moral nature. The answer to this objection could be that reputation provides the first motive to fulfil the promise, and therefore the first reason to expect fulfillment, but the obligation to honor the expectation provides an additional motive, and therefore an additional reason.9 But this answer runs straight into our circularity problem: the second reason cannot add significantly to the force of the first. If we have reason to doubt whether our partner is sufficiently sensitive to the motive of reputation, we should hesitate to rely on him, and therefore he will be unable to acquire an obligation. Either the non-moral reasons are inadequate to generate a supervenient obligation, or the obligation is sufficiently supported and hence redundant. 7 Hume, A Treatise of Human Nature, ed. L. A. Selby-Bigge, 2nd edition, revised by P. M. Nidditch (Oxford: Clarendon Press, 1978), III, iii. 5; cf. Lewis, supra, note 4, at 189; McNeilly, supra, note 4, at 72–73; Annette Baier, “Promises, Promises, Promises,” in Postures of the Mind (London: Methuen, 1985), pp. 188–193. On conditional strategies the modern classic is Robert Axelrod, The Evolution of Cooperation (New York: Basic Books, 1984). 8 Jon Elster, The Cement of Society (Cambridge: Cambridge University Press, 1989), pp. 117–118; cf. Michael Akehurst, A Modern Introduction to International Law (London: Allen & Unwin, fifth ed., 1984), pp. 8–11; Atiyah, supra, note 1, at 31, on modern credit economy with its blacklists of bad credit risks. 9 McNeilly, supra, note 4, at 77; MacMahon, supra, note 1. 356 GOVERT DEN HARTOGH In any case, the extension of the low-promising account cannot be the whole story, even if it is a large part of it. For promises are also occasionally given to people who probably will be unable to influence our reputation; and sometimes our interests in breaking the promise may be of greater weight than the loss of reputation involved. I do not want to suggest that it is never justified not to act on such “high” promises, but even then we do not believe them to be void. So the problem is not to be avoided. In the indirect account of promising the obligating force of the promise seems to be either redundant or – in the case of “high promises” – unexplained. 2. OPINIO IURIS A related problem of circularity can be found in international law. As a result of the intensification of international relations, explicit norm-creating procedures like treaties do not suffice to produce the required amount of binding rules. States therefore increasingly rely upon less formal ways of agreement and in recent cases10 the International Court of Justice has had to decide how these forms of agreement can be considered to be binding norms. In the first place the Court has stressed the importance of the principle of good faith as “one of the basic principles governing the creation and performance of legal obligations, whatever their source”.11 This suggests a kind of reliance-theory of obligation. More generally the traditional criterion of opinio iuris is given ever larger scope. It does not primarily concern the voluntary inducement of reliance, but a more general pattern of belief. For a legal norm to exist it is required in the first place that a certain convergence or regularity can be detected in the behavior of states – either a small amount of it, and no exceptions, or a great deal of it and only a few exceptions – but this is not sufficient; it must be accompanied by the belief that this behavior is required (or permitted) by a legal norm. For example, some particular rules can 10 E.g., Nicaragua Case, ICJ Rep. 1986, p. 14. Cf. P. P. Rijpkema, “Customary International law in the Nicaragua Case,” Netherlands Yearbook of International Law, 20 (1989), pp. 91–116. 11 Nuclear Test Cases, ICJ Rep. 1974, §49. A CONVENTIONALIST THEORY OF OBLIGATION 357 be found in almost all systems of private international law, but this is not enough to conclude that these rules are required by (public) international law, for it is not out of any conviction that they are required that these rules are incorporated into the municipal systems. In the early nineteenth century, France decided that French law should be decisive for the determination of the personal status (legitimacy, capacity to marry, etc.) of every person of French nationality, irrespective of his domicile. Nobody protested; hence it was tacitly admitted, and therefore true, that states are free to alter their rules of private international law at will.12 A classic example of the use of the criterion was provided by the Lotus case.13 A French naval officer, having caused an accident with fatal results for some Turkish sailors, was tried by a Turkish court. France disputed its jurisdiction. The Court considered that the precedents for the Turkish way of proceeding were very few in number indeed. However, on the one hand the few precedents which could be cited had never before met with opposition, and on the other hand no evidence existed that in the other cases the offenders had been handed over out of a sense of legal obligation.14 Apparently, no opinio juris existed, and hence no rule. The use of this criterion runs into exactly the same dilemma we met with the expectation-theories of promissory obligation.15 For the belief that the rule if binding must be either true or false. If it is 12 Akehurst, supra, note 8, at 49–50. 1927, PCIJ, series A no. 10. 14 Akehurst, supra, note 8, at 30–31. Cf. also North Sea Continental Shelf cases ICJ Rep. 1969. 15 John Finnis, Natural Law and Natural Rights (Oxford: Oxford University Press, 1980), pp. 238–245, attempts a solution by distinguishing the following judgments within the opinio: (a) it would be desirable to have a determinate, common and stable pattern of conduct determined by mutual expectations; (b) phi would be an appropriate pattern; (c) many states subscribe to (a) and (b) and converge in their practice to phi. These judgments warrant the conclusion that phi is a proper legal norm (imputes justified requirements), because of a valid meta-legal principle that it serves the common good (by solving a coordination problem) to conform to a pattern (like phi) which is the object of judgments like (a), (b) and (c). The weakness of this account is that it attempts to find a first independent reason for conformity, cf. below §3 on Lewis. Finnis does not recognize that the basic reason for conformity (the true meta-legal principle) depends on the same reason being recognized by the other parties as well, cf. below, at n. 46. 13 358 GOVERT DEN HARTOGH true, then the obligation already exists, and the opinion that it exists should be redundant as a criterion of its existence. But if it is false, the criterion seems to be discredited, for if no obligation exists, how can the false belief that it exists make it exist? The dilemma is felt acutely when a state wishes to alter existing rules. It cannot then avow this wish, for by doing so it would strengthen and not weaken the status of the incriminated rule. In such a case the new rule can only become law when states start breaking it, thereby “casting a vote in favor of its repeal”, and attempting, however disingenuously, to justify their conduct. When other states follow the example, or acquiesce in it being given, the attempt succeeds. International (or customary) law may seem a rather limited area of normative arrangements. But as a matter of fact the paradoxes of opinio iuris apply to every type of social rule, whether it involves an obligation or not. When are silver coins, beads, shells, pieces of printed paper, properly accepted as a means of exchange? When they are generally accepted. When is a particular locution grammatically correct? When it is generally agreed to be correct. When does a group of people form a nation? When they are conscious of forming one. When is something a work of art? Etc. Now I can explain why I believe Hart’s account of what it means for a social rule to exist to be defective. According to Hart a rule exists in a social group, if the members of the group act in a regular manner, expect each other to act in this way, notice deviations from the regularity, and expect each other to criticize and to correct deviating behavior.16 But a regularity of behavior can meet each one of these conditions and not be a social rule. For example, easily spoiling groceries are usually kept cool. Whoever puts old newspapers into his fridge, while putting his meat and milk on top of it, deviates from this regularity and leaves himself open to critical reactions.17 Why then is it improper to call this regularity a social rule? Because the critical reaction is independently justified in each individual case: it is simply stupid to allow food and drink to rot 16 H. L. A. Hart, The Concept of Law (Oxford: Oxford University Press, 1961), pp. 9–11, 54–56, 86–88, 96, 99–101; cf. Peter Winch, The Idea of a Social Science (London: Routledge, 1958), 28–33, 57–65; Raymond D. Gumb, Rule-Governed Linguistic Behavior (The Hague: Mouton, 1972), pp. 37–53. 17 Warnock, supra, note 1, at 45–47, with an equally parochial example: not taking an umbrella with you in British climate conditions. A CONVENTIONALIST THEORY OF OBLIGATION 359 unnecessarily. So not every “critical response to deviations” will do, only those which are dependent on other people having the same attitude. But if we have a proper social rule, why should I conform to the “regularity”, or having deviated, correct myself? It seems that you should be able to answer, “because it is the rule” (and not only, because it is the regularity).18 But then we have obviously moved in a circle. The rule exists whenever people believe it to exist; their belief is “collectively non-corrigible”.19 Clearly (pace Lagerspetz) this circle is a vicious one. For I would like to know what kind of belief this is: that a rule exists? If we try to explain it (“they believe that x, y, z”), do we have to say again at some point: and they believe all these things, because they believe that a rule exists? Then again, what is it that they believe? Even if, in our explanation, we have to refer to beliefs, as we probably must, these should not be beliefs about the existence of rules. I am not denying that Thomas’ axiom (what people believe to be real, is real in its consequences20 ) can often be used to explain social phenomena. But it can explain only their occurrence, not their essential nature. Institutional phenomenalism (esse est percipi) is a very queer doctrine. The so-called Scandinavian realists held that duty, law and money, could not possibly be things existing in the real world; they could exist only in the imagination of men. But it then becomes hard to explain what it is people falsely believe. It is only possible to believe in witches because there could be facts, other than the belief, making it true. 3. BUILDING HIGHER-ORDER EXPECTATIONS Hart’s original account of the existence of a social rule was deficient, as he came to recognize in his 1994 Postscript21 because it didn’t 18 As we will find Lewis suggesting, below §3. Eerik Lagerspetz, The Opposite Mirrors (Dordrecht: Kluwer, 1995), p. 14; cf. Govert den Hartogh, “Rehabilitating Legal Conventionalism,” Law and Philosophy 12 (1993), pp. 233–247. 20 W. I. Thomas, known from the famous chapter “The Self-fulfilling Prophecy” in Robert C. Merton, Social Theory and Social Structure (New York: Free Press, 1968). 21 H. L. A. Hart, supra, note 16, second ed. 1994, at 255–256. 19 360 GOVERT DEN HARTOGH acknowledge the fact that the actions of people following a rule have “a reference to each other” (Hume). That is precisely the point which is rectified by conventionalist theories of rules. However, I now want to suggest that such theories, at least in as far as they stay within a rational choice framework, are confronted with a problem of circularity all of their own.22 Consider David Lewis’s classical account. Lewis finds his explanatory context in problems of (more or less) pure coordination.23 He agrees with Schelling that such problems are generally solved – if they are solved at all – by converging on the “salient” alternative. (We have lost one another in Eaton’s and both decide to go to the bar in the upper store to meet each other.) Why is salience important? If one of the equivalent meeting-points (the bar, the exit, the customers’ service, and so on) is salient, this fact “indicates to me” that you will select it. Therefore, because the only thing that I am interested in is to meet you, I have a reason to select it as well. You will reproduce my reasoning, and because your preference pattern is identical to mine, the fact that I will go there will give you an additional reason to go there too. Mirroring each other’s reasoning, we will produce additional reasons ad libitum, until we are satisfied and, with a proper feeling of justification, proceed to the salient meeting-point. According to Lewis conventions operate in precisely the same manner: they rely on the fact that a meeting-point is made salient by precedent. I establish that there are two equivalent equilibria: everybody driving on the right and everybody driving on the left side of the road. Next I ascertain that until now one of those equilibria has constantly been the outcome of (almost) all choices of (almost) all parties involved; I conclude that this equilibrium is salient; and 22 Some conventionalist theories, in particular Robert Sugden’s, The Economics of Rights, Co-operation and Welfare (Oxford: Blackwell, 1986), escape the circle by relying on a mechanism of parametric learning by imitation of successful strategies. See also Bryan Skyrms, Evolution of the Social Contract (Cambridge: Cambridge University Press, 1996). 23 Lewis’ definition of such problems is unsatisfactory in several ways. I will interpret a 2-person game of pure coordination as a situation in which at least two proper equilibria (not necessarily coordination-equilibria) are present, both of which are also optima, and (more or less) of equal value to each of the parties. Cf. Govert den Hartogh, Wederkerige Verwachtingen (Amsterdam 1985), pp. 73–77; Id. Law as Convention (Dordrecht: Kluwer, forthcoming), ch. 1. A CONVENTIONALIST THEORY OF OBLIGATION 361 therefore that I may expect the others to make the choice which, with my additional choice, will bring all of us together at that equilibrium again; that I therefore have reason to make that particular choice; that the others will follow the same reasoning and therefore expect me to make that choice; that they themselves will therefore make that choice with even more confidence; that I therefore have that much more reason to do the same; and so on and so on. All those additional reasons are derived from the first independent one: the “indicating force” of the salient regularity. If the situation repeats itself the next day, then I must, in order to reconstruct the full reasons for my choice, begin again at that very beginning. Only this time the first indication for the choices of the others will be slightly strengthened once again.24 In each predicament, i.e., in every new round of the iterated coordination game, all parties involved build up their relevant practical reasoning from scratch. (To the extent, of course, that they account for their reasons: I am not concerned with actual thought processes, but only with the “order which is there”,25 with the adequate reasons for action which can be found in the circumstances.) But look at this first independent reason. Either it is strong enough to warrant me selecting a particular course of action. Then all the other additional reasons are strictly redundant. Or this first reason is in itself insufficient. Then the additional reasons have no solid foundation. The higher-order expectations therefore have no proper work to do; they rely on a prior belief which is either sufficient in itself or insufficient to carry them. To find a way out of the dilemma, let us first look at a situation which is not quite a coordination problem, because one of the players either slightly prefers one of the equilibria to the other(s), or has significant preferences over the non-equilibrium outcomes. Then the spiral of expectations is generated which ensures both players the success of their coordinating efforts. According to Lewis, the more levels of expectation the better.26 Each level adds a new reason, until the whole pile forms a sufficient one. Why 24 Lewis, supra, note 4, at 52–68; cf. Stephen R. Schiffer, Meaning (Oxford: Oxford University Press, 1972), pp. 149, 154. 25 G. E. M. Anscombe, Intention (Oxford: Blackwell, 1957), p. 80. 26 Lewis, supra, note 4, at 33. 362 GOVERT DEN HARTOGH don’t we keep forming higher-level expectations from here to eternity? Lewis answers, rather optimistically, that after the fourth or fifth step, the average windowless monad’s ability to abstract is too severely taxed. A more obvious answer would be that at a particular moment we have abundantly sufficient reasons. Enough is enough. Both replies rely on the assumption that we can indeed, at will, make the pile larger. Can we really do that? This must be a most curious faculty. From a finite amount of information being able to produce an infinite number of reasons. And, again, what is the point of the exercise? If the original reason was not sufficient, then the basis is too weak to make the inferences that lead to the extra reasons. But if the first reason is sufficient, then the rest are redundant. The whole procedure reminds one of the Baron Münchhausen’s trying to extract himself from the morass. Strictly speaking, it is not really true that these extra reasons can be wholly and completely derived from the original one. A number of ancillary premises have to be used in the reasoning process, premises concerning preferential orderings, knowledge (concerning preferential orderings, knowledge, etc.) as well as concerning the practical rationality and abstracting ability of the players. This is the scaffolding which must provide the edifice with additional support. It follows that, even if the original reason is as strong as a reason can possibly be, coordination cannot be guaranteed. It follows also that, if the original reason is not strong enough, coordination can succeed nevertheless. Take for instance our coordination problem in Eaton’s again. Suppose that you have a clear preference among the possible meeting points: this bar right at the top. I will then know where to find you right away (i.e., we start our mirroring thoughts from the independent basis). Suppose, however, that my belief in your alcoholic reputation is mistaken. I have no good reason to expect you in the bar. But you may know that I will impute the reason to you and so have a good second-order reason to go nevertheless. In a third version of the story I even know of your successful rehabilitation, but rightly believe you do not know I know. Obviously, we may go on to even higher levels. So, even though, to start with, I have no independent reason to expect you in the bar, on second thoughts I can replicate your assumptions, generating the expectation to find A CONVENTIONALIST THEORY OF OBLIGATION 363 you there nevertheless. Hence the rendezvous can be accomplished, even though no independently ascribed first reason ever existed. The spiral staircase of expectations only began to be ascended on the second, third or fourth floor. In this way coordination can start with correct and well-grounded higher-order expectations, even in the presence of false or misplaced lower-order ones. The correct lower-order expectations can still, on the way down, be formed, enabling us to make a coordinated choice. 4. “CONVENTIONS ARE LIKE FIRES” My dilemma arose from Lewis’ assumption that the rational reconstruction of the reasons for following the convention always has to start from a first independent reason to select the salient alternative. All higher-order expectations and the additional reasons they generate were to be derived from this basis. This assumption, however, is implausible. Take salience itself. It is not true that the aspect of salience “indicates to me” that you have an independent reason for your choice. If that were true, then I would have an independent reason myself as well. It is significant that in Lewis’ account “my” first reason already depends on the first independent reason of the others; for the outcome which is salient to me, and the fact that this outcome is salient to me, “indicates to you” that I will go there. But as a matter of fact salience gives neither of us any independent reason; I select the salient alternative because I expect you to do so, and I expect you to do so because I realize that you will expect me to do so. Even more, the relevant notion of salience in this context is itself a thoroughly interdependent one. I consider to be salient what I expect you to consider to be salient and vice versa. If our aim is to write down the same number, independently of each other, I am ill-advised to consider that 2011 has a special meaning for me. Salience in this sense cannot be an independent starting-point.27 27 Thomas C. Schelling, The Strategy of Conflict (Cambridge, Mass.: Harvard University Press, 1960), p. 94; cf. David Gauthier, “Coordination,” Dialogue 14 (1975), pp. 207–214; Margaret Gilbert, On Social Facts (London: Routledge, 1989), pp. 332–336; Robert Sugden, “Thinking as a Team,” Social Philosophy and Policy 10 (1993), pp. 77–82. Therefore the emergence of conventions cannot 364 GOVERT DEN HARTOGH The same applies to conventions. When we have regularly and successfully met one another at the same point, we can expect from one-another that in a subsequent analogical situation, our choices will lead us to the same meeting point. But if I would take the trouble to spell out my reasons, I could not start from any independent one, in order to build upon it the superstructure of higher-order expectations and additional reasons. The chain of my reasoning would not go downwards, but upwards. I don’t need to reconstruct the pattern of expectations by mirroring as a “windowless monad” the reasoning of the others, for I already know the pattern to exist. That is, a defining characteristic of being a member of a group is to participate in its mutual beliefs. Of course the belief is reinforced by the regular behavior which follows from it. But that doesn’t mean that the behavior indicates to us any independent reason on which the belief is constructed. We know that other road-users expect us to keep to the right; we have no reason to change this belief as long as they keep to the right themselves. We don’t need to know any independent reasons for their expectations. But if we do know their expectations, we know what to decide ourselves. Their expectations are a sufficient ground for our decision. And hence for their expectations! Everyone has reason to do what the other, for whatever reason, expects him to do. And precisely that fact is for everybody sufficient ground to continue in his expectations. Nobody needs more reasons. When a pattern of expectations has once surfaced, on whatever basis, then it is selfperpetuating without our ever having to return to that basis again. If this is correct, then it seems to me a reason not to identify (as Lewis does) convention with the regular behavior people display when repeatedly confronted by a coordination problem, but rather (as Shwayder did28 ) with the pattern of expectations that develops in such a situation. After all, the convention is what is quoted as the reason for particular conduct, and this, even in the final analysis, is not the constant mode of displayed behavior, but the pattern of expectations. be explained by the principle of salience. Choosing the salient alternative is itself a convention. 28 D. S. Shwayder, The Stratification of Behavior (New York: Humanities Press, 1965), pp. 252ff. A CONVENTIONALIST THEORY OF OBLIGATION 365 The power of patterns of expectation to chain people to a particular choice independently of any independent reasons is most apparent when a choice is no longer equivalent to rival alternatives, and therefore would not be chosen for independent reasons at all. Suppose the equilibrium a to be inferior to equilibrium b on the preferential scale of all players. Confronted with a choice between a and b for the first time all would choose b unhesitatingly. Yet a convention can exist for choosing a.29 In the Netherlands priority is given to traffic approaching a roundabout, although everyone can see that it would be more sensible for traffic on the roundabout to have priority. In such cases the conventional choice cannot be founded on replicated practical reasonings starting from independent “signifiers” for the anticipated behavior, for such signifiers point unambiguously to the b choice which is not made. Nothing other than the pattern of mutual expectations binds the players to the a choice; any suggestion of a dilemma is created by the convention itself. “These people are trapped.”30 Consider Kavka’s analysis of the case of “perfect tyranny”.31 According to Kavka it is possible for a regime to be built on nothing but fear, if every member of the population (including the Tonton Macoute or Securitate agents) believe: • either (a) there will be at least someone prepared to obey the dictator from a sense of loyalty, or a motive of gain; • or else (b) there will be at least someone who will believe (a); • or else (c) there will be at least someone who will believe (b); etc. It is not necessary, Kavka tells us, for a first punisher to exist, or for someone to believe him to exist. It is not even necessary for anybody 29 Lewis concludes that in order to have a coordination problem “different coordination equilibria do not have to be equally good – only good enough so that everyone is ready to do his part if the others do.” Lewis, supra, note 4, at 50. But without a convention there would be no coordination problem at all. 30 Id., at 98. For other examples of patterns of expectation which hold us captive, see Thomas C. Schelling, Micromotives and Macrobehavior (New York: Norton, 1978), pp. 31–32 (sending Christmas and New Year cards), cf. Elster, supra, note 8, at 138–149. 31 Gregory S. Kavka, “Rule by Fear,” Noûs 17 (1983), pp. 601–620. Or the analogous case of the millionaire, Lewis, supra, note 4, at 27. (Both situations are assurance games rather than proper coordination problems, cf. n. 38.) 366 GOVERT DEN HARTOGH to identify the level of his relevant higher-order beliefs. But I want to go even further than that. If a pattern of mutual expectations has been established, we may all know quite well that no “first punisher” exists at all, know each other to know it etc. Nevertheless we may know that most of the others will be prepared to help the dictator to execute his punishments, because they know that most of the others are prepared to do so. None of us has to trace his expectations downwards. The lower-order expectations are all of them derived from the higher-order ones, not vice versa. This is especially clear if we don’t accept Kavka’s “no loop assumption”, as seems plausible. It may well be the case that I am afraid that someone will punish me because he is afraid that someone will punish me because . . . etc. because he is afraid that I will punish him because I am afraid . . . (Again we do not need to identify the place-markers. What matters is that I have no reason to exclude myself from any of the indefinite classes in the further chain.) I am back at my own expectations at the original level, and so I can go on without ever reaching a bottom-level at all. “Conventions are like fires: under favorable conditions a sufficient concentration of heat spreads and perpetuates itself. The nature of the heat does not depend on the original source of heat. Matches may be our best fire starters, but that is no reason to think of fires started otherwise as any the less fires.”32 The “original source” which Lewis is thinking of here is agreement (mutual “low promising”); yet, habit, precedent and salience33 are nothing more than “matches”. Let me retrace my steps. I began by showing that higher-order expectations are not redundant when an independent justification for lower-order expectations is absent. I went on to argue that a transparent pattern of mutual expectations, having been constructed on an independent basis in a number of successive analogous predicaments of choice, may start to live its own life. People have such expectations, know each other to have them, and so on, without having to ask for any further justification of their beliefs again. 32 Lewis, supra, note 4, at 88. And orders: the prudent player sticks to the strategy of his team etsi Gullit non daretur. This is the basic mistake in overestimating the role of legislative authority in law. 33 A CONVENTIONALIST THEORY OF OBLIGATION 367 Taking both insights together, we can assert the following. Once the pattern of expectations is self-supporting, it is the higherorder expectations which justify the lower-order ones, and not vice versa.34 5. EXTENDING THE SOLUTION I now want to extend this solution of the third of my circularity problems to the first two. This cannot be done in a straightforward way. The context in which the obligations of “high” promises and of international (or any customary) law arise, obviously is not captured by the notion of a game of pure coordination.35 In such a game no obligations are needed, because once a meeting-point has become conventional, or salient in any other way, no one has any incentive to choose to go anywhere else. And therefore the pattern of mutual expectation of convergent behavior is self-reinforcing. The way “low” promising functions may, as I suggested before, be properly analyzed in this way. But in the case of “high” promising the fact that you expect me to act in a way conformable to your interest, does not give me a non-moral reason to do so. Perhaps it is precisely your expectation and the way you act on it which gives me my real chance of gaining at your expense. This may well be the very reason why a high or obligating promise was needed in the first place. One way of dealing with this problem has been suggested by Postema and by Lagerspetz, two pioneers of the conventionalist theory of law.36 It may be the case that, on encountering situations with, e.g., the structure of a Prisoner’s Dilemma, people develop 34 The Nash equilibrium as a solution concept in game theory confronts us with similar problems. If a game has a unique solution, it must be an equilibrium, but why assume that each game has a unique solution? See Robert Sugden, “Rational Bargaining” in Michael Bacharach and Susan Hurley (eds.), Foundations of Decision Theory (Cambridge, Mass.: Blackwell, 1991), pp. 294–315. There is no independent reason for each player to play an equilibrium strategy and hence to expect the other player to do so. 35 Leslie Green, The Authority of the State (Oxford: Oxford University Press, 1988), ch. 4; E. Ullmann-Margalit, “Is Law a Coordinative Authority?” Israel Law Review 16 (1983), pp. 350–355. 36 Lagerspetz, supra, note 20, at 198–207; Gerald J. Postema, “Bentham on the Public Character of Law,” Utilitarianism 1 (1989), pp. 52–56. 368 GOVERT DEN HARTOGH new preferences vis-à-vis such situations, for instance a preference for fair dealing. (They could be altruists as well, but then we would rather have to say, counterfactually, that what would have been a P.D., given the non-tuistic preferences of the players, really is not, taking into account all relevant preferences. A preference for fairness, on the other hand, is a response to situations with a possible free-riding option; the reference to the P.D. character of the situation enters into the proper description of the preference.37 ) Such a moral preference would transform the nature of the game from a P.D. into a coordination problem,38 and the latter would be solved by a pattern of mutual expectations in the usual way. This solution is on the right track, but it is not quite satisfactory. What it does not recognize is the fact that trustworthiness and fairness are responses, not only to P.D. situations, but to patterns of mutual expectations arising in them; the reference to the pattern enters into the proper description of the preferences as well. Being trustworthy is being prepared to honor the expectations of cooperation of the other players, and being fair is refusing to exploit them.39 The exercise of cooperative virtues does not simply involve making cooperative choices in situations in which pure forward-looking rationality would tend to lead to defective ones; it presupposes that people expect each other to make cooperative choices. If I am a fair and trustworthy person, it is not only relevant to me that by making the cooperative choice I expect you to make, you will give your fate into my hands as a matter of fact, but also that you will do so knowingly. What I take into account, is not only your vulnerability, but also your trust. 37 This answers the standard-objection that an appeal to moral preferences to solve the P.D. is inadmissible since the numerical utility values are supposed to reflect all relevant preferences. R. Duncan Luce and Howard Raiffa, Games and Decisions (New York: Wiley, 1957), p. 96; Thomas C. Schelling. “Game Theory and the Study of Ethical Systems,” Journal of Conflict Resolution 12 (1968), p. 42; Ken Binmore, Game Theory and the Social Contract, Volume 1: Playing Fair (Cambridge, Mass.: MIT Press, 1993), pp. 95–98. 38 Or rather into an assurance or “staghunt” game, characterized by an optimal but risky and a sub-optimal but safe equilibrium. In this game a pattern of mutual expectations is also self-reinforcing. 39 And justice is the constans et perpetua voluntas ius suum cuique tribuere, “ius suum” referring to a pattern of expectations arising in bargaining games. A CONVENTIONALIST THEORY OF OBLIGATION 369 But if this is true, it seems we have a new problem of circularity on our hands. Cooperative virtues are exercised in response to other people’s expectations, but these expectations themselves are only formed on the basis of ascribing those virtues. How does cooperative virtue succeed in playing its part? Suppose we are in a one-shot two-person P.D., we have to make our choices simultaneously, and you are known to be a fair and trustworthy person. Given the relevant preferences which gave the situation its P.D. structure to begin with, I can still only expect you to defect, and you know it. Doesn’t it make any difference that you are trustworthy? Only if I am trusting you and you know that I am trusting you. Can’t I then “just go ahead and expect”?40 There are two problems with this suggestion. In the first place the fact that having a belief will introduce a causal process making the belief true, is not a justification for having that belief. It is possible that you believe that you will have a pleasant afternoon, this brings you into a good mood, and so you really have a pleasant afternoon. But it is not possible rationally to start entertaining the belief because you expect this to happen. That is like taking a pill for its placebo effect. As H. A. Prichard once put it, “unfortunately, being neither lunatics nor pragmatists, we should recognize that any attempt to attain the position in this way would fail, since our knowledge that if we all had this belief our having it would be extremely useful to ourselves would not in the least lead us to have it”.41 Incidentally, it is for similar reasons that I believe that all volitional theories of promising are doomed to failure. They all exhort the agent to “go ahead and intend”. But intention can only be formed in reflection of pre-existing reasons for action, you cannot create those reasons by intending.42 Would it not be possible for me to bring you to cooperation by signalizing my trust? Perhaps it would, but that would be a subtle form of deception. Compare the following case. If you believe that I expect you to pay for my drink, you will do so. Knowing this, I 40 Narveson, supra, note 1, at 223. Prichard, supra, note 5, at, 176. Cf. Annette Baier, “Trust and Antitrust,” Ethics 96 (1986), p. 244. 42 Govert den Hartogh, “Cooperative Intentions and their Autonomous Effects,” unpublished. 41 370 GOVERT DEN HARTOGH signalize to you that I am expecting you to pay. As a result, I may be right in expecting you to pay, but my action is not justified by my expectation. It’s a trick. Even if it would be possible to decide to go ahead and expect, because of the autonomous effects of the decision, a second problem with the suggestion remains. In the situation I described there is not sufficient temporal space to establish a causal link. You cannot respond to my expectation, you have to anticipate it, so it is impossible for me to create a fait accompli. But how can you anticipate my “decision”, knowing that it will be made itself in anticipation of your appropriate response? Our beliefs concerning each other’s choices are, even if our cooperative dispositions are shining like a jewel to both of us, fully dependent on each other. I can only expect you to exercise your trustworthiness, if you can expect me to start expecting, and vice versa. As there is no independent starting-point for any such expectation, none of us can break the spell. It will by now be clear how I propose to deal with this problem. If a pattern of mutual expectations of cooperative behavior has somehow emerged, cooperatively disposed people will have reasons to honor the expectations, and will thereby confirm and hence stabilize the existing pattern. No independent starting-point of justifying the expectations exists, and none is needed, because the lowest-order expectations are derived from the higher-order ones. I expect you to cooperate, because I know that you know I expect this, and you know that I know that you are a trustworthy person.43 My circularity-problems all have the same general form. I believe that you have a reason to do x, and the fact that I believe so is itself the reason, or an essential part of the reason, for you to do so. You don’t have a belief-independent reason, so I don’t have a belief-independent reason for my belief. The solution of the problem in the case of conventions was to distinguish between my actual particular first-order belief, and the standing pattern of self-reinforcing mutual beliefs of indefinite order, and to take the first to be justified by the latter, not vice versa. It is generally mutually and transparently believed that in circumstances C people will do x; you are in circumstances C (driving), therefore you have every reason to do x, and I am justified in believ43 If no pattern of expectations exists, the problem is unsolvable. A CONVENTIONALIST THEORY OF OBLIGATION 371 ing you to have this reason. Analogously, it is generally mutually and transparently believed that in circumstances C people will do x; and some people are sometimes disposed to let this very fact count as a moral reason to do x: they are to this extent trustworthy. If I can identify the trustworthy people and ascertain the extent of their trustworthiness, I am justified in expecting them to do x. The pattern of expectations is not built up on independent reasons of the people involved, where such reasons are constituted by nonmutual beliefs and preferences (including, perhaps, preferences for fair and loyal cooperation). If the pattern exists, it generates its own reasons, including “new” preferences, and thereby reinforces and stabilizes itself.44 6. OBLIGATIONS AND MUTUAL BELIEFS Consider opinio iuris again. Why should the United States refrain from laying mines in Nicaraguan coastal waters? Because a pattern of mutual expectations exists among nations that such acts will be refrained from, if no state of war is declared. So why are nations justified in expecting this? You shouldn’t say at this point: because each of them has an independent reason for refraining from such acts, namely, that they are wrong. In that case the opinio would be redundant indeed. But neither should you say: if they have no independent reason, the expectations are unjustified, and the opinio is mistaken. They expect each other to omit such acts, and for guileless actors this fact is itself one good reason to do so. Even if it is in the United States’ (short term) interest to lay those mines, a trustworthy member of the community of nations should not do such things. Nicaragua lets its fishing fleet go out of its harbors, trusting they are not in danger because the country is not in war with any other nation. In that case it would be a breach of faith to lay mines. And therefore we can say that Nicaragua is justified in its expectations. Its first-order expectations are warranted by the common pattern of higher-order ones (and not vice versa). 44 Conventionalist theories usually treat the questions of the explanation of the emergence and of the stability of norms as almost identical. If my amendments are accepted, these questions require completely different answers. 372 GOVERT DEN HARTOGH In this explication of the notion of a social rule, self-reference is scrupulously avoided. Practical mutual beliefs are mutual expectations, and the object of the lowest-order expectation is simply a type of action: everyone in S expects that everyone does x. This, of course, conforms to Lewis’ analysis. So where does the appearance of circularity come from? Whence the strong inclination to believe that we explain something about the existence of the rule, if we account for our critical response to deviations by saying, “You should do x, because it is the rule”? Lewis’ analysis is right. The lowest-order expectation simply refers to behavior. But Lewis is mistaken in his conception of the order of justification. Lowest-order expectations depend on higher-order ones, not conversely; otherwise the higher-order ones would be redundant. But to make this point, we need not introduce any circularity into our analysis. The set of circumstances C which “trigger” the actual investment of expectations, obviously may include someone’s voluntary acts, including strictly conventional ones which have no meaning except this triggering. So we find our non-circular account of high promising. A standing transparent pattern of mutual expectations exists that everybody who, under circumstances C, says “I promise to do x at time t” (crosses his heart, nods, etc.) will do x at time t. (This is the reality referred to by the mystical concept of the so-called “constitutive rule”.) If I am a trustworthy person, I will feel bound, having said these words, to honor the expectations I thereby knowingly induced in you. And therefore you are justified in (first-order) expecting me to do x at time t. This is really a direct theory of the obligatory force of promising, but it incorporates the valid elements of the indirect theory. What the promise intentionally produces directly is a higher-order expectation that the promiser will do x.45 This he can only do because of the existence as a matter of contingent fact of a pattern of mutual expectations (a “practice”). From the higher-order expectation a first-order one may be derived. The derivation appeals to a principle of trustfulness (or fairness). A first-order expectation justified in this way by higher-order expectations and a moral principle, I propose to call an obligation. To this extent the adherents of the indirect theory 45 And not an internal constraint on the will, as in the volitional theory. A CONVENTIONALIST THEORY OF OBLIGATION 373 were right, the existence of the obligation depends on a principle of keeping faith. But what they didn’t see is that the successful inducement of a (lowest-order) expectation is dependent on this “principle” as well. What you intend to induce in high promising is reliance on your trustworthiness. The mistake involved is equally common in direct theories of promising. A practice is described in terms of its “constitutive” rules, and then it is said: and a valid moral principle (autonomy, fidelity, fairness, or whatever) exists, requiring you to conform to the rules.46 One fails to observe in this way that the practice could not exist, unless we could count on each other generally to honour this principle. We cannot describe the “institution”, and then leave it to the autonomous individual to make up his moral mind; for the mutual ascription of a moral motive is the foundation of the institution. This invites an obvious objection. It seems to follow that only a person who is trustworthy, or even known to be trustworthy, can assume an obligation. The idea is not quite as absurd as it might seem. Suppose you live in a very small community, all members of which know each other from frequent interaction. In such a community the person who habitually lies will eventually be unable to make any deceiving statements whatsoever. For people will simply stop believing him. In the same way the unreliable person will be unable to execute an intention to have other people rely on him. But in larger communities there will be many transactions in which no sufficient evidence concerning the individual parties will be available. In these cases we cannot trust them on account of their personal characteristics, but only on account of the general trustworthiness of the people in this community. Therefore the obligation does not rest on the promisor as the particular individual she is, but only as a member of a particular community. It is this anonymous pattern of mutual expectations which I propose as the proper analysans of the concept of a social norm. 46 Cf. John Rawls, A Theory of Justice (Cambridge, Mass.: Harvard University Press, 1971), pp. 344–350; Charles Fried, Contract as Promise, A Theory of Contractual Obligation (Cambridge, Mass.: Harvard University Press, 1981), pp. 16–17; J. H. Nieuwenhuis, Drie Beginselen van Contractenrecht (Deventer: Kluwer, 1979); Finnis, supra, note 15. 374 GOVERT DEN HARTOGH The climax of this process of anonymization – the stage of the legal contract – is reached when a person, personally known to be unreliable, succeeds in making use of the anonymous norm to make himself responsible for his future conduct nevertheless. This may, significantly, require him to use a more formal means of proceeding than a more reliable person would need. 7. SOME FINAL COMMENTS I have argued that contractual and customary obligations are derived from transparent patterns of mutual expectations. This covers the whole field of obligations: voluntary and involuntary ones. Of course, I am using the concept of “obligation” in a narrow sense, and not in the wide Kantian sense in which it is equivalent to any moral requirement whatsoever.47 Duty and obligation in the narrow sense are only part of the moral domain, which already appears from the fact that recognizing either presupposes both a conception of the good and some aspects of moral character. The notions of duty and obligation on the one hand and of (cooperative) virtue on the other mutually imply each other. For that reason it makes no sense to set the ethics of virtue and the ethics of duty in opposition against each other. By realizing this it is also possible to answer the objection that on my conception murder, rape etc. are only morally wrong because people mutually expect each other to abstain from such acts. Note that in my extended use of the notion of convention the connotation of arbitrariness is absent. In particular when a transparent pattern of mutual expectations which has an obligating force, its content will not be arbitrary, for the forms of mutual adjustment of actions they establish is not a solution to a coordination problem. In a Prisoner’s Dilemma for instance there is only one optimal outcome, and it can be identified as such without any reference to a convention. Similarly the existence of the relevant expectations concerning murder, rape etc., is explained by the fact that these things are mala per se, and our primary moral reason for abstaining from them is not that 47 Cf. Bernard Williams, Ethics and the Limits of Philosophy (London: Fontana Press/Collins, 1985), ch. 10. A CONVENTIONALIST THEORY OF OBLIGATION 375 they are violations of duty. It seems to me a virtue of my theory that it allows us to say this. In positivist theories of law it is often denied that legal obligation is a form of moral obligation.48 On my account it is. It is true that legal obligations largely derive from the exercise of legal (legislating and adjudicating) authority. But that doesn’t mean that the conventionalist analysis is not applicable. As a matter of fact, there is a strong analogy between legal and promissory obligations, promise being an exercise of authority over oneself. Not surprisingly then, the explanation of (practical, not epistemic) authority confronts us with the familiar circularity problem. There has been some discussion of the question what kind of authority has analytic priority: de facto or de jure. Take for instance the authority of a referee. Clearly it can only be reasonable to acquiesce in his decisions if they are also recognized by the other players. Practical authority functions only in virtue of the interdependence of individual responses to its exercise. Therefore, someone is a legitimate authority only if he is a de facto one. But on the other hand it is analytically true that a de facto authority is someone generally believed to be a legitimate one. Again, the belief seems to be “collectively non-corrigible”. The solution, again, is that someone is an authority in a group, if the members of the group mutually expect each other to follow his decisions, irrespective (within limits) of their content. What they expect is conforming behavior; but this lowest-order expectation is justified by being derived from higher-order ones. Analogously to “low promising” we will find forms of “low authority”, people by and large expecting each other to conform because it is individually utility-maximizing for them to do so. But often they can only expect each other to conform on the basis of the mutual ascription of a cooperative virtue. This might be fairness, or trustworthiness, but perhaps the most relevant virtue in this context is to be prepared to accept a reasonable compromise, e.g. in a bargaining game concerning the specification of a package of collective provisions.49 48 E.g., H. L. A. Hart, Essays on Bentham (Oxford: Oxford University Press, 1982), p. 266. 49 Govert den Hartogh, “Authority and the Balance of Reasons”, in Robert Alexy und Ralf Dreier (eds.), Rechtssystem und Praktische Vernunft, Band I, ARSP-Beiheft 51 (Stuttgart: Franz Steiner Verlag, 1993), pp. 136–144, substantially revised in Den Hartogh, supra, note 23 (forthcoming), ch. 7. 376 GOVERT DEN HARTOGH In this case what the law at least purports to create, is the obligation to incur certain costs, the obligation being justifiable by reference to the standing pattern of higher-order mutual expectations to obey legal authority. Department of Philosophy University of Amsterdam Nieuwe Doelenstraat 15 1012 CP Amsterdam The Netherlands hartogh@philo.uva.nl View publication stats