THE METHOD OF A TRULY NORMATIVE LEGAL SCIENCE Jaap Hage1 Universities of Maastricht and Hasselt e-mail: jaap.hage@maastrichtuniversity.nl Summary This paper contains an argument to the effect that the proper method for legal science depends on what one takes to be the nature of science, the nature of the law and the kind of questions that are addressed in legal science. It starts from three assumptions, namely that: a. science is the collaborative pursuit of knowledge, b. the law consists of those norms which ought to be enforced by collective means; c. the proper standard to determine what ought to be done is what maximises the long term happiness of all sentient beings (the H-standard). On the basis of these assumptions the following positions are argued: 1. Legal science, in the sense of a description of the law, is not impossible for the reason that it is a normative science; 2. In abstract the method of all sciences, including legal science, is to create a coherent set of positions that encompasses ‘everything’, and therefore also beliefs about the law. 3. The proper method for a normative legal science consists primarily of the methods of sociology, psychology and economics, because the ultimate question to be answered is the collective enforcement of which norms satisfies the H-standard. The more traditional hermeneutic methods only play a role to the extent that they establish positive law that contributes to happiness by providing legal certainty. 1 The author wants to thank Anne Ruth Mackor, Jan Smits, Michal Araszkiewicz and the (other) participants of the Tilburg workshop on the methodology of legal research for useful comments on a draft version of this paper and discussions on the nature of legal knowledge and science. In this paper I will outline a method for a truly normative legal science. With ‘truly normative’ I mean that this legal science provides the answer to some version of the question ‘What should we do?’. I will argue that the issue of the method for legal science hangs together with views on the nature of science, on the nature of law, on the justification of alleged knowledge, and – as I will argue – in the end with almost everything. The second section of this paper will be devoted to an argument that the method for a normative science is essentially the same as that of a science that deals with ‘facts’. In this first section I discuss a number of assumptions that are needed to get the argument started. The precise status of these assumptions will be clarified in section three, which deals with the method of normative legal science in particular. Section four summarises the argument of this paper. a. The nature of science If we want to know what the proper method for legal science is, we should at least have some idea of what we mean by ‘science’.2 Science has to do with the pursuit and accumulation of knowledge.3 Moreover, it aims to systematise this knowledge. How this systematisation takes shape depends on the object of the knowledge. In case of historical sciences the system derives from the way in which facts and events explain each other. In the case of physical sciences, the system consists in the laws that are formulated and that are used to explain and predict events and facts, and in the way in which laws are derived from each other. In mathematics the system consists in the axiomatisation of a subdomain and in the derivation of theorems from these axioms. A third characteristic of science, which explains important other characteristics, is that science is a social phenomenon. It is impossible to be the only scientist in a field, at least in the long 2 However, the editor of Blackwell’s ‘A Companion to the Philosophy of Science’ (Oxford, 2000), WH Newton-Smith, refused in the Introduction to this book to give a definition, because the prospect to succeed would be bleak. 3 According to AR Mackor, ‘Explanatory non-normative legal doctrine’, in this volume, an important task of sciences, including legal doctrine, is to explain legal norms. This view is well compatible with the view of science exposed here, because the explanation Mackor is after boils down to the knowledge that particular legal norms can be derived from (amongst others) other norms. A similar view is defended by Van Rhee (C.H. van Rhee, ‘Geen rechtsgeleerdheid, maar rechtswetenschap!’, in Rechtsgeleerd Magazijn Themis 2004, p 196-201) , who argues for a legal science in which the coherence between legal rules and legal principles is made explicit. run. Science is a cooperative enterprise aimed at the acquisition, accumulation and systematisation of knowledge. The advantage of science over individual acquisition of knowledge is that scientists can build on the results of their colleagues. To quote Newton: ‘If I have seen further it is only by standing on the shoulders of giants.’ 4 Let us assume that science is a way in which people collaborate in the pursuit and systematisation of knowledge. If such collaboration is to be possible, several conditions must be met. First it must be assumed that the aspired knowledge is, at least approximately, the same for everybody involved in the cooperation. If everybody would have his own ‘truth’ it would be impossible for one person to build on the results of other persons.5 This demand would, in the eyes of many, exclude aesthetics and astrology from the arena of science. Very often the assumption that truth is the same for everybody is made on basis of another assumption, namely that knowledge describes a world which is mind-independent and therefore the same for everybody (ontological realism). A true description of this independent world would be the same for everybody too. It is possible, however, to assume a truth that is the same for everybody without endorsing ontological realism. Mathematical truth would, according to many, be a case in point. b. Science and method A second precondition for the possibility of cooperative knowledge pursuit is that there exists at least to a large extent agreement on what count as good reasons for adopting or rejecting a potential piece of knowledge.6 Here is where method comes into the picture. For what is a scientific method? 4 Original Latin: ‘Pigmaei gigantum humeris impositi plusquam ipsi gigantes vident.’ Letter to Robert Hooke of February 15th 1676. Source: http://en.wikiquote.org/wiki/Isaac_Newton (last consulted on December 15th 2009). 5 As Mackor pointed out to me, it is possible that there are several subgroups within a scientific community with different ‘truths’ in the sense of points of belief convergence. Every such group could theoretically have its own science. 6 Arguably, there are other preconditions for science in the sense of collaborative knowledge acquisition. One may think in this connection of ways in which scientific results are published, financed etc. For the present purposes the two mentioned preconditions, a shared knowledge object and a common method, are the most important ones. In one sense of the word, it is a way of going about in doing science. It is a kind of procedure that is to be followed if the results are to count as ‘scientific’.7 An example of such a procedure would be the empirical cycle as described by De Groot8, or the Herculean method described by Dworkin in the chapter ‘Hard Cases’ from Taking Rights Seriously.9 In another sense, a scientific method indicates what count as good reasons for adopting or rejecting a potential piece of knowledge. Take for instance the mathematical thesis known as the Goldbach conjecture10, that all even numbers bigger than two can be written as the sum of two prime numbers. One mathematician would count on proof to establish the truth of this thesis, while another mathematician would take a large collection of random even numbers, check whether they can be written as the sum of two primes, and decide that from this sample it is clear that the conjecture is almost certainly true . If they consider their own method as the only legitimate one, these two mathematicians cannot cooperate in the pursuit of knowledge on number theory. The adoption of a particular method in this second sense boils down to agreement on what count as such good reasons. Since such an agreement is a precondition for science as collaborative knowledge acquisition, a shared method is almost by definition a precondition for science.11 Reasons in general, and therefore also reasons for accepting or rejecting a particular piece of potential knowledge, are facts that are relevant for what they are reasons for or against.12 The adoption of a method is a choice for what counts as relevant. It is also a choice concerning the kind of data that must be collected in order to argue for or against a potential piece of knowledge. For instance, on a hermeneutic method for legal science, the relevant data for a particular legal conclusion might be that this conclusion is supported by the literal interpretation of a statute, which is adopted as an authoritative text. Therefore a legal 7 Scientific method, even in the sense of a way of going about in doing research, is general. This means that a method is not an algorithm to be followed in a particular research project. 8 AD de Groot, Methodologie, many editions, ’s-Gravenhage, Mouton, chapter 1. 9 See R Dworkin, Taking Rights Seriously, 2nd ed London, Duckworth 1978. 10 See http://en.wikipedia.org/wiki/Goldbach%27s_conjecture. (Last consulted on November 2nd 2009). 11 That science requires a shared method does not exclude that this method is mostly implicit, or that it changes in the course of time. If such a change is drastic, for instance if physics comes to be based on experiments rather than on interpretation of authoritative texts, the nature of the science changes too. 12 An extensive discussion of the nature of reasons can be found in JC Hage, Reasoning with Rules (Dordrecht, Kluwer, 1997) chapter 2. See also C Redondo, Reasons for Action and the Law (Dordrecht, Kluwer, 1999). researcher should consult this text, and apply, possibly amongst others, a literal interpretation to it.13 The proper way of going about in legal research, method in the first sense, is to a large extent14 determined by method in the second sense of the recognition of particular kinds of data as relevant for the issue at stake. It is this second sense of ‘method’ that will be at stake in the rest of this paper. Science in the sense of collaborative knowledge acquisition is practically impossible without such a method. c. Method and the object of knowledge The idea of a method is often connected to disciplines, such as law, physics, mathematics, biology, medicine, history, sociology, or psychology. In the following I will continue to write about the methods of a discipline, but this is, in a strict sense, incorrect. Which facts count as reasons for or against a conclusion depends on the type of conclusion and therefore on the research question at issue. One discipline may deal with several kinds of research questions and then different methods are relevant in answering these questions. Legal science is a case in point. The question what the criminal law of a jurisdiction is - the traditional doctrinal question - differs, for instance, from the question how the contents of the criminal law developed in the course of time – the legal historical question. It is improbable that the same kinds of facts would be relevant to answer these two questions. So, if within a discipline different kinds of research questions are being asked, the issue of method should be focused on a type of research question, rather than on the discipline as a whole.15 For the following discussion of the method of legal science, I will focus on the description of the (contents of the) law of a particular jurisdiction at a particular place and time.16 13 It is in this connection that Van Hoecke mentions normative and authoritative sources such as legislation and case law, as empirical data used in legal doctrine. (‘Legal Doctrine: Which Method for What Kind of Discipline?’, in this volume). 14 Other factors might be conventions on how results are to be published, and restrictions on how research may be financed. 15 A consequence of this position is that a researcher should be explicit on the kind of research question that (s)he tries to answer, and in particular on the impact which this has for the choice of a method. Especially where different questions within one field require different methods, clarity about the kind of question that is addressed is crucial. 16 The clause ‘of a particular jurisdiction’ will be relativised in section 3. The methods of a scientific discipline are normally chosen because the participants in the discipline assume that these methods lead to the kind of knowledge pursued in their discipline. A good example is formal logic. One of the questions with which formal logic deals is what the theorems of a particular logical system are. Logicians believe that this question has one correct answer. Moreover, each potential theorem either is or is not a theorem of the system at issue. Logicians cooperate in identifying the valid theorems and by giving reasons (proofs) why the proposed theorems are valid. Moreover, the alleged theorems and the accompanying proofs are published, to share the results with other logicians who can build upon them, and who are also enabled to check whether the alleged theorems have been proven. Logicians consider proofs to be relevant because they assume that proofs lead to conclusions which are true, not only for the person who gave the proof, but also for all other logicians. In fact, they even attempt to prove that proofs lead to true results by showing that a particular proof theory is ‘sound’. There exists an independent test, in the shape of model theoretic semantics, which determines whether a particular theorem is true, and a particular logical calculus is sound (a recommendable characteristic) if its proofs lead to theorems that are true according to the semantics.17 The point of this example is that scientific disciplines tend to assume that there is truth to be had and also that the methods they employ are normally suitable to discover this truth. Formal logicians assume that proofs lead to true theorems; theorists of the physical sciences assume that the cycle of hypothesis formulation, empirical testing of hypotheses, and improving the hypotheses on the basis of the test results, leads to ever better (in the sense of more true) theories18, and moral philosophers assume that mutual adaption of concrete moral intuitions and general moral principles lead to ever better moral theories.19 17 More on the nature of logic and in particular the relation between proof theory and (model theoretic) semantics can be found in thorough introductions to formal logic, including S Haack, Philosophy of Logics (Cambridge University Press, 1978) and LTF Gamut, Logic, Language, and Meaning (2 volumes) (Chicago, University of Chicago Press, 1990). 18 Cf KR Popper, ‘Truth, Rationality, and the Growth of Scientific Knowledge’, in his Conjectures and Refutation, 4th ed (London, Routledge and Kegan Paul, 1972) pp 215-250. More on the idea of verisimilitude in C Brink, ‘Verisimilitude’, in Newton-Smith, Companion to the Philosophy of Science, pp 561-563. 19 The standard reference here is J Rawls, A Theory of Justice, 1st ed (Cambridge, Harvard University Press, 1971) section 9. See also N Daniels, ‘Reflective Equilibrium’, http://plato.stanford.edu/entries/reflectiveequilibrium (last consulted on October 11, 2009). Those moral philosophers who consider themselves as being involved in a normative scientific enterprise will also assume that the better theories are closer approximations of a moral ‘truth’ which should be accepted by The methods of scientific disciplines are often based on implicit theories concerning the nature of the discipline’s objects and the suitability of these methods for obtaining knowledge about objects with that nature. Because mathematical theorems are different from physical laws, it takes different data to argue for the truth of theorems than for the existence of physical laws. Changing insights into the nature of a discipline’s knowledge object may lead to changes in method. If, for instance, the law is not (anymore) considered to be an answer to the question what to do, but rather a body of rules, rights and principles that happen to exist at a particular time and place, we might stop arguing about the contents of the law by pointing out the consequences of particular rules, and revert to the study and interpretation of authoritative texts or the behaviour of leading jurists. A discipline and its methods are part of a wider body of (hypothetical) knowledge, which includes views on the nature of the discipline’s knowledge objects and theories on how and why particular data are relevant to establish knowledge about such objects. In connection with the proper method of legal science this would mean that the view concerning this method hangs together with a view on the nature of the law, and a view on which data are relevant to determine the truth – if there is any to be had – of potential pieces of legal knowledge. At this point I want to mention the possibility that legal ‘science’ does not aim at the pursuit of knowledge about something at all. Many lawyers are involved in keeping the law of a particular jurisdiction in good shape. This is done by describing the law as it is, incorporating recent changes caused by, for instance, new legislation and case law, into the body of legal knowledge, by evaluating the existing law, and by proposing changes to it or even- if one is the position to do so - by bringing about the desired changes.20 This is an important task of legal ‘science’, and it is benefited by an academic level of dealing with the law, but it is not science in the sense of the word used here of cooperative knowledge acquisition. It is rather a form of highly qualified maintenance of the legal system. There may be some overlap in method with ‘real’ legal science, but maintenance of the legal system is a different discipline from legal science, and I will not deal with it here. everybody, whether they agree or not. Mackor pointed out to me that moral philosophers might, like mathematicians, confine themselves to axiomatizing a body of moral rules. That is right, but then these moral philosophers would not be engaged in a normative enterprise anymore. 20 See also the description of the practice of the legal discipline by P Westerman, ‘Open or autonomous? The debate on legal methodology as a reflection on the debate on law’, in this volume. d. Three views on the nature of law In most disciplines a method reflects a view of the discipline’s knowledge domain. For instance, in physics it used to be assumed that nature obeys certain ‘laws’ and that these laws manifest themselves in observable phenomena. Observations can be used to induce hypotheses about the laws and to test the laws through predictions of new observations.21 The proper method is therefore to use observations to induce laws from (empiricism), or to falsify predictions (critical rationalism). In mathematics the idea is that the domain consists of a set of theorems that somehow ‘follow’ from the discipline’s axioms. The appropriate method is then to deduce the theorems from the axioms (or to prove – if that is possible - that a potential theorem cannot be proven). Similarly, one would expect the method of legal science to reflect a view about the law.22 PURELY PROCEDURAL LAW It is possible to distinguish at least three fundamentally different views on the nature of the law. One view is that questions about the contents of the law – even ‘easy’ questions - have no true answers and that the law consists merely of a set of acceptable argument forms, such as an appeal to legislation, to case law, to legal principles, human rights, legal doctrine, and the standard canons for legal interpretation and legal reasoning. Legal argument is not aimed at finding the contents of the law, because there is no such a thing. It is aimed at convincing one’s auditorium of a particular legal position. Some arguments are more authoritative than others23 and should therefore be more convincing, but what counts in the end is not whether 21 See for instance R Harré, The Philosophies of Science (Oxford University Press, 1985) chapter 2 and AF Chalmers, What is this thing called Science?, 3rd ed (Indianapolis, Hacket ,1999) chapters 4 and 5. 22 That is what one should expect, but as yet I know only a few authors writing on legal method, who based their views on a theory about the nature of law. One example is C Smith, ‘Het normatieve karakter van de rechtswetenschap. Recht als oordeel’ (The normative nature of legal doctrine. Law as judgement), forthcoming in Rechtsfilosofie en Rechtstheorie 2009-3), who bases himself on the purely procedural view of the law that is described below. Another example is Mackor, Explanatory non-normative legal doctrine, who bases her explanatory account of legal method on the Law as Social Fact view. 23 If the authoritativeness of arguments is measured by standards which exist a a matter of social fact, the procedural view of the law merges with the law as social fact view. The ‘only’ difference between the two would then be that the procedural view of the law focuses on arguments by means of which legal positions can be supported or attacked, while the law as social fact view focuses on the positions that turn out to be right given the argument forms that are commonly considered to be authoritative. This difference is comparable to that between traditional and dialogical presentations of logical systems. Cf EM Barth and ECW Krabbe, From Axiom to Dialogue, (Berlin, Walter de Gruyter, 1982). the correct position was defended – because the correct position does not exist – but which argument was most convincing in the sense of effective. The law would be, to use Rawls’ phrase, purely procedural24, with the not unimportant clause that the procedures that constitute the law, the acceptable argument forms and the materials to which they refer – legislation, treaties, case law, and custom - to a large extent constrain the possible outcomes.25 The law for a concrete case, of for a case type, would be the outcome of a battle of arguments.26 Legal science in the sense of collaborative knowledge acquisition requires the possibility of agreement. If the law is purely procedural, this possibility can only exist if the nature of the legal procedure constrains the possible outcomes of a battle of arguments to such an extent that only one outcome is viable.27 But then the law is not purely procedural anymore, because most legal discussions would have only one possible outcome if played by the procedural rules, and then the contents of the law are fixed, more or less in the same way as mathematical theorems are fixed by the axioms and the rules of the proof system. If the law is purely procedural, however, that is if the procedural rules in combination with the contents of the legal sources do not determine the outcomes of legal argument battles, there is no basis for agreement on the contents of the law28 and legal science in the sense of collaborative pursuit of knowledge is impossible. LAW AS SOCIAL FACT A second view of the law holds that the law exists as a matter of social fact, independent of what individuals may believe about it, but dependent on what sufficiently many sufficiently 24 A Theory of Justice, section 14. 25 These constrained options correspond to the norm-contentions described in Mackor, Explanatory nonnormative legal doctrine. 26 This view is defended by Smith, Het normatieve karakter van de rechtswetenschap, by A Soeteman, Rechtsgeleerde waarheid, valedictory address, Vrije Universiteit June 19th 2009, and by J Smits, Omstreden rechtswetenschap (BJu, Den Haag, 2009), p 93f. The latter author also gives references to other Dutch adherents of the view that the law is purely procedural (discursive). Formal elaborations of the procedural view of the law are discussed in my paper ‘Dialectics in Artificial Intelligence and Law’ in JC Hage, Studies in Legal Logic (Dordrecht, Springer, 2005) pp 227-264. 27 Soeteman, Rechtsgeleerde waarheid, p 15, writes about legal truth, and therefore seems to assume that the law (often) sufficiently constrains legal arguments to make one outcome the right one. Whether this single right answer can be identified easily or authoritatively is a different matter. 28 There may be agreement on what the law is not, however. In that sense, some knowledge is possible even on a purely procedural view of the law. important members of a social group think about the contents of the law and think about what others think about it.29 A special variant of this view is that of law as institutional fact, according to which most of the law exists thanks to rules that specify what counts as law.30 This view of law as social fact has two advantages. First it explains why the law appears to be a matter of fact, independent of what individual persons think of it, and that the contents of law depend on a particular jurisdiction. And second it explains why lawyers tend to argue about the law as if it already exists and as if two conflicting legal positions cannot both be true. The law as social fact view has also an important disadvantage, namely that there would be less law than seems at first sight. If law exists as a matter of social fact, there cannot be more law than is fixed by social reality. In particular there cannot be law about which knowledgeable lawyers fundamentally disagree.31 And yet, such disagreements seem to occur frequently. Does this mean that these lawyers do not disagree about the law, but rather about how the law should be expanded to cover the case at issue? And if this is the case, why do these lawyers not clearly separate the two discussions, one about the law that actually exists and the other about the most desirable way to create new law? Somehow the arguments about the contents of the law and what would be desirable legal solutions for types of cases seem to conflate. Is this a matter of methodological confusion, of pious deceit, or is something else the case? 29 One version of this view was made popular in HLA Hart, The Concept of Law, 2nd ed (Oxford University Press, 1994). 30 This approach is described in, amongst others, N MacCormick and O Weinberger, An Institutional Theory of Law (Dordrecht, Reidel, 1986), E Lagerspetz, The Opposite Mirrors (Dordrecht, Kluwer, 1995), DWP Ruiter, Institutional Legal Facts and Legal Institutions (Dordrecht, Kluwer, 1993 and 2001), and N MacCormick, Institutions of Law (Oxford University Press, 2007). In this volume, the institutional approach is explicitly adopted in J Bell, ‘Distinctiveness of legal research and the implications for comparative law’. 31 At first sight, this drastic conclusion seems avoidable by allowing arguments of commonly accepted types on the basis of commonly accepted legal sources. But then there are three possibilities: 1. these arguments do not lead to an unique conclusion; 2. there is an unique conclusion, but this is not commonly accepted; 3. there is a unique conclusion, which is not commonly accepted. Only in the last case we can speak of law as social fact. The second case (as the first) would be a variant of the procedural view of law. THE NORMATIVE VIEW OF THE LAW The third view of the law assumes that something else is the case. According to this view, the law is essentially an answer to the question what to do, and more in particular what to do by means of rules32 which should be enforced collectively, usually by means of state organs.33 Notice that according to this third view, the law is, not what is actually enforced collectively, but what ought to be enforced collectively. To state it in an overly simplified way: the law is an ought, not an is.34 Therefore I will call this the normative view of the law. On this normative view there is principally no difference between the law as it is, and the law as it should be.35 Moreover, the law would be a branch of morality, if morality is taken as that set of standards that indicates what would be good and right things to do all things considered and taking the interests of all human (sentient) beings into account.36 The obvious advantage of the normative view is that it explains how discussions on the contents of the law often deal with what is desirable. The equally obvious disadvantage is that it seemingly fails to explain how the law is the same for everybody, why law appears to be primarily national law, and how the law is related to such matters of fact as the contents of legislation and of case law, judicial decisions, and social practices such as the canons for interpretation and legal reasoning. This disadvantage does not need to be real, however. The ‘positive’ law, which exists as a matter of social fact, is an important - in fact by far the most important - factor that determines the law in the sense of rules that should collectively be enforced. It is highly desirable that the law can function properly in regulating human society 32 The idea that the law consists of the rules that should be enforced collectively does not involve the other idea that this should apply to rules on an individual basis. One might well opt for the version that complete bodies of rules should be judged on whether they should be enforced collectively, with the proviso that many rules belonging to such a body, in particular the procedural rules (Cf. JH Merryman, The Civil Law Tradition, 2nd ed, Stanford University Press, Stanford, 1985, p 70/1 ), do not lend themselves well to enforcement. 33 At least in theory it is possible to have a different normative view of the law, a view according to which the law is an answer to the question what to do, but not the question which rules to enforce by collective means. It might for instance be the question which rules serve the general interest. I will not pursue this alternative normative view of the law here any further. 34 This is oversimplified because it assumes that an ought is not an is. In ‘What is a norm?’ in my Studies in Legal Logic, pp 159-202, I have argued why the proper distinction is between rules and facts, and not between is and ought. I take these two distinctions to be quite different from each other. 35 For the famous contention otherwise ‘The existence of law is one thing; its merit or demerit is another’, see J Austin, The Province of Jurisprudence Determined, (several editions), note to the 5th lecture. 36 Clearly, other circumscriptions of morality are possible. and for that purpose it needs to be stable, the same for everybody, and easily recognizable.37 In practice this means that the law must by and large be positive law.38 The difference with the law as social fact view is, however, that positive law is ‘real’ law (‘real law’ in the sense of the law that should be enforced by collective means) not because that is the social practice, but because, and to the extent that, the positive law ought to be enforced collectively. Moreover - and this has immediate implications for the method of legal science – the positive law is only ‘real’ law to the extent that it contributes to the recognisability of law and to legal certainty. This means that if the ‘positive’ law can only be established by means of some contestable interpretation, it cannot fulfil its coordinating function anymore and loses its presumptive force as law. In this paper I intend to outline a method for legal science as a description of existing law39, on the assumption that the normative view of the law is correct. Legal science would then be a normative science, aiming at the collective pursuit and systematisation of normative knowledge, in particular knowledge which rules should (here and now) be enforced collectively. This view of legal science has some similarities with, but should nevertheless be distinguished from the view, promoted by Smits, that legal science is normative in the sense that it deals with the question what the law should be.40 Although Smits is not very explicit about the nature of the law41, it seems that he considers the law to be a set of rules etc. that exist in social practice. Legal science should, according to Smits, indicate what this practice should be. In my opinion, the ‘real’ law, as distinguished from the merely positive law, is itself an answer to a normative question, and legal science as description of this ‘real’ law, aims at providing this answer. Despite this difference, the view of Smits on the nature of legal science 37 For these and other ‘internal’ demands on the law, see LLH Fuller, The Morality of Law, revised edition, (New Haven, Yale University Press, 1969) ch 2. 38 This position is far from new; it can already be found in Aquinas. See his Summa Theologica I-II, Qu 95. 39 I therefore ignore other kinds of legal science, such as the explanation of existing law, or the comparison of the law from different jurisdictions. 40 Smits, Omstreden rechtswetenschap, 70. 41 However, he does describe the law as a spontaneous order and a product of natural selection (Omstreden rechtswetenschap, 79 and 81). has an important similarity to my view, because we both assume that legal science deals by and large with the question which rules we should have, or enforce by collective means. The method that I apply to formulate the proper method for legal science is to argue why the proposed legal method contributes to the pursuit of knowledge about which norms should be enforced collectively. In particular it is not an analysis of the method used in contemporary legal doctrine. This method is, as is correctly pointed out by Van Hoecke42, essentially hermeneutic. From the fact that a hermeneutic method is in fact used in the development of legal doctrine, it does not follow that this is the proper method, however. In this sense, I am not a methodological naturalist.43 As I will argue in section 2e, the method that is actually used in doctrinal legal science is no more than a starting point in the process of deciding which methodological positions stands up to critical scrutiny. a. Why normative science seems problematic It is a popular view that normative science is not well possible.44 The reason is generally some form of non-cognitivism concerning normative (and evaluative) issues. It is customary to distinguish between the realms of is and ought and to be an ontological realist with respect of the realm of is, and to be a non-realist with respect to the ought. With regarding to is-matters, there would be a mind-independent reality which is the same for everybody45 and which makes every factual proposition true or false.46 With regard to ought-matters, such a mind42 Van Hoecke, Legal Doctrine. 43 On methodological naturalism in legal theory, see B Leiter, Naturalizing Jurisprudence (Oxford University Press, 2007), pp 30-46. More in general about methodological naturalism is H Kornbluth (ed), Naturalizing Epistemology, 2nd ed (Cambrdige, MIT Press, 1994). 44 For the Netherlands this position was taken by GE Langemeijer, Inleiding tot de studie van de wijsbegeerte des rechts (Zwolle, Tjeenk Willink 1956) p 296 and Gerrit de Geest, ‘Hoe maken we van de rechtswetenschap een volwaardige wetenschap’ in NJb 2004/2, pp 58-66 (implicit). According to Smits a normative legal science is possible, but in his view it is not possible that such a science would lead to consensus (Omstreden rechtswetenschap, p 111). Given my charaterisation of science, this last view of Smits implies the negation of his first view about the possibility of a normative legal science. 45 For a discussion of this form of realism, see Michael Devitt, Realism and Truth, 2nd ed (Oxford, Blackwell, 1991) chapter 2. 46 For the purpose of this paper, I ignore the complications of vague propositions and of sentences with nondenoting terms in referring positions. These issues are to be dealt with by, for instance, respectively fuzzy set theory and fuzzy logic (see: http://en.wikipedia.org/wiki/Fuzzy_logic; last consulted on October 12th 2009) , and theories on the relation between sentences and the referring expressions in it. See for instance AC Grayling, An Introduction to Philosophical Logic, 3rd ed (Oxford, Blackwell, 1997), chapter 4. independent reality would be lacking. To state it bluntly: whether we agree about it or not, there would be a true answer to every question of fact, while there is no such true answer concerning normative questions. What ought to be done would not be a matter of facts that are the same for everybody, but a matter of taste, or of choice, which may have a different outcome for different persons, even if they are all fully rational. There is no common ground which can function as a foundation for agreement and where there is no ground for agreement, so runs the argument, there is no room for science. The same issue can also be approached from a logical point of view.47 To justify an oughtconclusion by means of a deductively valid argument, at least one of the premises must be an ought-sentence too. Moreover, for a successful justification, the premises of the justificatory argument must be either true, if they are factual, or justified. Since presumably normative premises cannot be true, they must be justified, but this justification requires an argument with at least one normative premise, which must be justified … etc. However long the justificatory argument chain is made, it never touches solid ground in the form of premises which have all truth values. This line of argument has become so familiar that attempts to base normative conclusions on solely factual premises are discarded without much discussion as committing the ‘naturalistic fallacy’.48 I will argue that this familiar line of argument against the possibility of a normative science either is much weaker than is usually assumed, or hits purely ‘factual’ science just as hard as it hits normative science. The central piece of my argument is a theory about the nature of justification. b. Positions There are many different things which can be justified, such as beliefs, actions, decisions, verdicts, etc. At first impression one might think that these different objects of justification require different forms of justification, but this impression is only to a limited degree correct. All forms of justification can be reduced to variants on justification of behaviour (including forbearance). This is obvious for actions, and since decisions and verdicts can be brought under the category of actions (taking a decision, or giving a verdict with this particular content), it should be obvious for decisions and verdicts too. The same counts for using rules. 47 For an example, see RM Hare, The Language of Morals (Oxford, Oxford University Press, 1952) chapter 2. 48 The best exposition of this point is to my knowledge still PW Taylor, Normative Discourse (Englewood Cliffs, Prentice Hall, 1961) chapter 9. It is somewhat less obvious for beliefs, but the justification of a belief with a particular content can be interpreted as the justification of accepting this belief content. Accepting something can, for justificatory purposes, be treated as a kind of mental action. And just as it is possible to accept belief contents, it is possible to accept goals, values and principles. It is even possible to continue along this line, by treating the justification of the different forms of actions as the justification of accepting ‘that these actions are the ones that should be performed (under the circumstances)’.49 In this way, all forms of justification can be treated as the justification of accepting ‘something’. As a catch-all term for things that can be mentally accepted, I will from now on use the word ‘position’. Building on this definition, I will use the expression ‘position set’ for the set of all positions accepted by a person. c. Local and global justification In the literature on legal justification, justification has sometimes been pictured as a deductively valid argument.50 In such an argument the conclusion (what is justified) must be true given the truth of the premises. The idea behind this kind of justification is that the ‘justifiedness’ of the premises is transferred to the conclusion, analogous to the way in which the truth of the premises is transferred to the conclusion in more traditionally conceived deductive arguments. It seems to me that this picture is mistaken in at least two ways. First, because it suggests that ‘being justified’ is a characteristic of positions that is similar to truth, only somewhat ‘weaker’. Second, because it overlooks the essentially global nature of justification. In a deductively valid argument, the conclusion must be true if the premises are true. This means that the truth of the conclusion is guaranteed by the truth of the premises, and that nothing else is relevant for this truth.51 For instance, if the statements ‘All thieves are punishable’ and ‘Jane is a thief’ are both true, the statement ‘Jane is punishable’ must be true too. Whatever 49 See Giovanni Sartor, Legal Reasoning. A Cognitive Approach to the Law (Dordrecht, Springer, 2005) chapter 3. 50 See for instance R Alexy, Theorie der juristischen Argumentation, (Frankfurt, Suhrkamp, 1978) pp 273-278 (on ‘internal’ justification), and Neil MacCormick, Legal Reasoning and Legal Theory, on ‘first order justification’ (Oxford, Clarendon Press, 1978) chapter 2. 51 This should be read as ‘irrelevant from an argumentation-technical point of view’. The only thing that is really relevant for the truth of a conclusion is whether this conclusion corresponds with the facts. But that has nothing to do with the argument from which the conclusion follows. else may be the case, this cannot influence the truth of the statement ‘Jane is punishable’, unless it has the implication that one of the premises is false after all.52 For instance, the fact that Jane is only five years old either has no impact on Jane’s punishability, because the statement ‘All thieves are punishable’ is still considered to be true, or – which is more plausible – it has impact, because it makes the statement ‘All thieves are punishable’ false. The general point here is that the truth of the conclusion of a sound justificatory argument (deductively valid with true premises) is not influenced by additional information. If the premises of a deductive justificatory argument are true, its conclusion must be true, and then probably also justified, whatever else is the case. Therefore, it is not necessary to consider additional information, because this information cannot change the conclusion anymore. Deductive justification is local in the sense that it needs only consider the premises and the conclusion of a deductively valid argument. The reader who thinks this is unrealistic is probably right. Not because deductively valid justification is not local, but because deductively valid justificatory arguments are seldom offered. They are seldom offered, because they require premises the truth of which cannot be established. Take our example about Jane. It requires the premise that all thieves are punishable. That premise can only be established if it is already known that Jane is punishable (that would be a necessary condition), or if there is a rule that makes all thieves punishable, without exceptions. The former demand would beg the question, because we need the premise about all thieves to justify a belief about Jane. The second demand is unrealistic, because rules tend to have exceptions, and these exceptions cannot be enumerated. Real life justification is normally based on premises that support the conclusion without guaranteeing its truth. If Jane is a thief, this is a reason to believe that she is punishable, but there may be other reasons which invalidate this conclusion. One such a reason would be that Jane is only five years old. In general a position is justified if the reasons pleading for acceptance outweigh the reasons against accepting it to a sufficient degree. But this means that all reasons for or against acceptance must be balanced and that means in turn that justification must be global. A position which is justified in the light of a particular set of 52 This may be interpreted as a reason why justification on the deductive account of it is global too. But then the global nature does not lie in the deductively valid argument itself, but in the justification of the premises. other positions need not be justified in the light of an even larger set of other positions, because this larger set may contain additional reasons against adopting it.53 In logic there is a technical term for a similar phenomenon: nonmonotonicity. A logic is nonmonotonic if a conclusion that follows from a set of premises does not need to follow from a wider set of premises.54 Analogously we can say that justification is nonmonotonic because a position that is justified in the light of a set of other positions, needs not be justified in the light of a still wider set of positions.55 Nonmonotonicity and the global nature of justification go hand in hand. The ‘normal’ justification of a position is always relative to a particular position set. To get rid of this relativity, one needs to idealise and to assume that it is possible to consider a position in the light of all other positions. A position would be justified in an absolute sense if it is justified in the light of all positions. Obviously, absolute justification is an unrealistic notion for practical purposes, but it is useful as a tool for thinking about the nature of justification. d. Integrated coherentism Theories of justification are prone to be criticized for suffering under what Hans Albert has dubbed the ‘Münchhausen-trilemma’, after the famous baron who tried to pull himself out of the morass by his hairs.56 Because the premises of a justificatory argument would need to be justified themselves, there seem to be only three possibilities: a. some premises (e.g. those resulting from sensory perception under ideal circumstances) are dogmatically accepted as true or justified; b. the need to justify the premises leads to an infinite regress, because the arguments used to justify the premises also use premises which need to be justified, and so on …; c. the premises of a justificatory argument are indirectly justified by the conclusion of the justificatory argument; in other words the justification would be circular. 53 In fact, it is even more complicated, because apart from reasons against a position, there may be reasons why the reasons for adoption are in general not reasons after all, or are excluded in this special case. More details on the ‘logic of reasons’ in Reasoning with Rules, chapter IV, and in Studies in Legal Logic, chapter 3. 54 See, for instance, ML Ginsberg (ed), Readings in Nonmonotonic Reasoning (Los Altos, Morgan Kaufman, 1987) pp 1-2. 55 A more extensive argument to the same effect can be found in the paper ‘Law and Defeasibility’ in my Studies in Legal Logic, pp 7-32. 56 Hans Albert, Traktat über kritische Vernunft (Tübingen, Mohr, 1980) section 2. Let us assume that Albert’s analysis is correct and that these are the only three possibilities. The question then is whether this is problematic. My answer would be that it is not, because all justification is relative to the set of everything one accepts, one’s position set. Justification is necessarily circular in the sense that the justification of every position that a person accepts is based on this person’s position set. On what else could it be based? Not on reality itself, because our contact with reality is through what we believe about reality.57 The global nature of justification forces us to adopt a coherence theory of justification. The idea behind coherentism is that the justification of a position consists in the position being an element of a wider set of positions which somehow ‘cohere’ with each other.58 Coherentism has the advantage over foundationalism, its main competitor in the theories of justification, that it does not require a foundational set of positions which are considered to be justified without further reasons, and which are therefore made immune against criticism.59 This advantage comes with at least two seeming disadvantages. One alleged disadvantage is that if justification is considered to exist in a mutual relation between positions only, the connection with the ‘world outside’ seems to be lost. Haack writes in this connection about the consistent fairy story objection, or - even more picturesque – the drunken sailors argument, because the elements of a coherent theory may keep each other upright like drunken sailors who cannot stand by themselves.60 I will address this seeming disadvantage in section 2e. The second disadvantage is that the notion of coherence is hard to specify. What does it mean if a set of positions is said to cohere? There are very complicated accounts of this notion of coherence61, but it seems that a simple account is possible. If coherence is treated as a 57 Perceptive states that lead to beliefs are not relevant for justification, even if they can explain some of our beliefs. It is those beliefs (positions) that play a role in justification. 58 This is one possible version of coherentism. For a brief overview of several alternatives, see K Lehrer, ‘Coherentism’, in J Dancy and E Sosa (eds), A Companion to Epistemology (Oxford, Blackwell, 1992) pp 6770. 59 If there are reasons for immunising some positions against criticism, these very reasons are the proof that the privileged positions are not privileged at all, but derive their special position in the set of all positions from their relation to other positions in the same set (the reasons). This kind of ‘immunisation’ is well compatible with coherentism. For details, see my Studies in Legal Logic, p 42. 60 S Haack, Evidence and Inquiry (Oxford, Blackwell 1993) pp 26-27. 61 An example would be the theory defended by Peczenik and Alexy. See R Alexy and A Peczenik, ‘The Concept of Coherence and its Significance for Discursive Rationality’, in Ratio Juris 3 (1990), pp 130-147. characteristic of a set of positions that includes not only beliefs, but also all kinds of standards, a set of positions may also contain the standards that are used to determine whether a particular position is justified in the light of a set of other positions. In fact, a comprehensive position set would contain such standards. That makes it possible to use the position set in the definition of coherence. A somewhat simplified account would be the following: A position set is coherent if and only if it includes every position that should be accepted in the light of its content (the counterpart of logical closure), and does not include any position that should be rejected according to its own content (the counterpart of consistency). Because this notion of coherence refers to standards that are contained in the coherent set itself, I have called it ‘integrated coherentism’.62 When I write about coherence in the rest of this paper, I mean integrated coherence. Since a coherent position set includes everything that should be accepted according to itself, such a set will presumably be infinitely large. For realistic justification we will have to work with more limited sets, under the assumption that the limited set is a representative part of a coherent infinite set. I will call this the soundness assumption. This soundness assumption is defeasible, in the sense that it may turn out to be incorrect in the light of new information. Suppose for instance that we are dealing with the belief that Jane is punishable. We know that Jane is a thief and that the rule exists (is valid) that thieves are punishable. In the light of this limited position set, we should also accept that Jane is punishable, and the position set should be expanded accordingly. The soundness assumption here includes that there are no other reasons relevant for the punishability of Jane. When the soundness assumption has been shown63 to be wrong, the finite position set will have to be changed into another set for which the soundness assumption has not been refuted (yet). If it has been shown that Jane is five years old and that minority (in the sense of criminal law) takes one’s punishability away, the soundness assumption has been shown to be wrong. The position set must be expanded to make it include that Jane is five years old and that being a minor takes one’s punishability away. And given this expansion of the position set, the belief that Jane is punishable will have to be retracted from it. 62 JC Hage, ‘Law and Coherence’, in Ratio Juris 17 (2004), pp 87-105. See also Studies in Legal Logic, pp 5859. 63 Notice the procedural nature of this demand. It does not require truth, nor justifiability; it requires an actual change in a position set. The relevancy of this dynamic aspect is discussed in JC Hage, ‘Dialectics in Artificial Intelligence and Law’. e. Spontaneous positions A familiar objection against coherence theories is that a coherent position set may be isolated from reality. A set of positions may be coherent while all positions contained in it are false. The elements in the set justify each other, but there is no guarantee that the content of the set as a whole somehow reflects reality. This would be problematic, because a position set will normally include beliefs about the ‘world outside’. The limited set consisting of the beliefs that Jane is a thief and that Jane is punishable and the rule that thieves are punishable may be coherent, but does it really justify the belief that Jane is punishable? Maybe Jane does not even exist! From the coherence of the set nothing seems to follow about the truth of the beliefs contained in it. Can such an isolated set justify these beliefs? This is a familiar objection, but on closer inspection it is not very strong. To see why, one needs to consider how a coherency test of justification will operate in practice.64 One does not come up with a coherent set of positions from scratch. Normally one starts from an already existing set.65 The contents of a real position set, that is a set that is entertained by some real person, will have two kinds of determinants, rational ones and a-rational ones. The rational determinants make that an existing set is corrected – new positions are added and existing ones are removed – because rationality requires this given the rest of the set. (Remember that the demands of rationality are also specified by the position set.) The a-rational determinants cause ‘spontaneous’ changes to the contents of the set. New positions are added as a consequence – notice the causal terminology - of perception, memory, intuition, or any other factors which cause what a person accepts. A person may, for instance, accept something because he mistakenly believes that this is rational in the light of what else he believes. 64 There is also a very brief refutation of the objection, namely that it confuses truth and justification. That a position is untrue is no objection against a position being justified, or – better – against a person being justified in accepting this position. This would be different if a person knows, or should have known, that a belief is untrue. But then the problem is not the falsity of the belief, but the acceptance of the belief that an accepted position is false, which amounts to inconsistency of the position set. 65 Raz writes in this connection about the ‘base’. See J Raz, ‘The Relevance of Coherence’, in J Raz, Ethics in the Public Domain (Oxford, Clarendon Press, 1994), pp 277-325. This base may also include methodological guidelines. In the case of doctrinal law it may, for instance, contain the guideline that legal doctrine is to be developed by means of understanding authoritative texts (the hermeneutical method). If I am right, this guideline should rationally be replaced by a method that better reflects the nature of law and of science. Existing positions are removed, because they are forgotten or abandoned for irrational or arational reasons. These irrational or a-rational influences on a position set are relevant because position sets are biased toward the past. Whether a new position should, from the rational perspective, be added, or an existing one removed, depends on the present contents of the set. To see why, one should notice that a particular position can have one of three statuses in the light of (the rest of) a position set: 1. it should be adopted (if not already present) because this is rational; the position is acceptable; 2. it should be removed (if it is already present) because this is rational; the position is rejectable; 3. it is neutral in the sense that it should neither be adopted nor removed; the position is suspendable. As long as a position set is not coherent – that means in practice: always66 - the judgment whether a particular position should be added or removed should rationally not be made on the basis of a full position set, but only on the basis of the acceptable positions and of the suspendable positions included in the set.67 Positions that should not be in the set themselves, should not play a role in determining what else should be in the set. Moreover, suspendable positions remain in the set, even if there is no reason to adopt them, and codetermine what is acceptable. So suspendable elements can be justified because they belong to a coherent position set, without being justified by particular reasons in the set. This may, for instance, hold for beliefs caused by perception, or for evaluative judgments caused by ‘intuition’. As long as there is no reason to reject them, they will be maintained. Moreover, neutral elements can also play a role in determining whether other elements can remain in the set and are therefore justified. And finally, since rational modifications can only take place on the basis of an existing position set, the original elements of any realistic position set must have been spontaneous positions, that is neutral elements. This is one part of the argument why real position sets do not ‘hang in the air’. Real sets stem from sets of spontaneous positions, and there is where we should look for the desired ‘contact with reality’. 66 A real position set will always remain incoherent, if only because adoption of everything that should rationally be adopted leads to an infinitely large set, while the human mind is finite. 67 Acceptable elements, which were not actually accepted yet, should also be taken into account. A particular position is justified relative to a position set if it is an element of this set, and if this set is coherent. Moreover, given the important role of spontaneous positions, the position set should be a real one, held by a particular person.68 So, the relativity of justification does not only concern position sets, but also persons. A position is justified relative to the set of positions held by a particular person. Therefore it is better to speak of a person being justified in accepting something69, than of a position being justified. f. The outside world How can we be sure that spontaneous positions reflect ‘the world outside’? We cannot, simply because we cannot compare positions with the world. We can only compare them with what we accept about the world, including spontaneous positions. Somehow, the ‘outside world’ seems out of reach, and the objection that a coherent theory may have no contact with reality seems on the one hand to be correct, but on the other hand also to be unavoidable. Yet, this is not the whole story. Of some of our positions we believe that they reflect an external world. We believe that there is such an external world to begin with, and we also believe that this world influences our spontaneous positions. It should be noted, however, that these beliefs, like all of our other beliefs, belong to our position set, and can be corrected on the basis of the rest of the set. The existence of an external world is a bit like the existence of so-called ‘theoretical entities’. Entities like electrons cannot be perceived, but their existence is derived from other things that we can perceive. With the external world it is a little different: we believe that we can perceive it, but its existence is still theoretical in the sense that we postulate its existence on the basis of perceptual impressions. The world is, so to speak, necessary to explain our perceptions of it.70 68 The idea that position sets may be held by collectivities, for instance in the form of the ‘body of scientific knowledge’, or by a group of experts, is important, but beyond the scope of this paper. See KR Popper, ‘Epistemology Without a Knowing Subject’ and ‘On the Theory of the Objective Mind’, in KR Popper, Objective Knowledge (Oxford, Clarendon Press, 1972) pp 106-190. 69 This what is called ‘doxastic justification’, which is opposed to so-called ‘propositional justification’ (a particular position is justified). See R Neta and D Prichard (eds)¸ Arguing about Knowledge (London, Routledge, 2009) p 151. 70 This should not be read as stating that we believe the external world to exist only because that would explain our beliefs about it. Our cognitive apparatus is such that we spontaneously believe that (most of) our perceptive impressions are impressions of the external world. Eg we do not only spontaneously believe that the sun is shining, but also that the sun is shining in the ‘outside’ world. In fact, the latter assumption is so natural that we automatically take the first belief to be identical to the second. In this connection, agreement plays an important role. If different persons have the same beliefs, this may be taken as a reason to assume that there exists an independent object of belief that causes the unanimity of the beliefs about this object. For instance, the ‘objective’ existence of a table explains why we all see the table and believe that it is there. Agreement in beliefs needs not be explained by an objective external world, however. A second type of explanation would be that agreement on a position is the outflow of the functioning of our cognitive apparatus. Think in this connection of mathematics. Mathematicians tend to agree on many results of their science, but only some of them attribute this agreement to an objectively existing world of mathematical objects, such as for instance natural numbers. Another example would be that our moral intuitions are, at least to some extent, innate.71 A third possible explanation of agreement is that a position is the result of a procedure that is designed in such a way that it leads to the same outcome for (almost) everybody. A playful example would be the procedure to throw 10,000 times with an unbiased dice, which leads for almost everybody to the outcome of more than 1000 times a six. A legally more relevant example would be that legal arguments based on the same rules and cases and using the same canons of interpretation and argumentation lead in easy cases to the same outcomes for almost everybody.72 From the fact that agreement on a particular position may be a sign that this position reflects an outside world that is the same for everybody, it does not follow that where agreement is lacking a position does not reflect the outside world. For instance, we believe that the position ‘There is water on the moon’ reflects the outside world, but there was (at the moment of writing the first draft of this paper) no agreement yet amongst scientists whether this position is true. However, because we assume that the position reflects the outside world, we tend to However, we are able to question this spontaneous belief, and if it is questioned a reason for adopting it (again) is that the external world explains both the existence of our spontaneous beliefs about it, and the convergence of (some of) our beliefs about it with the corresponding beliefs of other persons. 71 This position is argued by Hauser. See MD Hauser, Moral minds: How nature designed our universal sense of right and wrong (London, Abacus, 2008). 72 This may even be so by definition, if easy cases are defined as those cases which lead to agreement amongst those who argue by the rules of law. This position is argued in JC Hage, R Leenes and A Lodder, ‘Hard cases: a procedural approach’, in Artificial Intelligence and Law 2 (1994), pp 113-167. believe that an increase in relevant knowledge should, in the end, lead to agreement.73 Where we do not even expect that an increase in knowledge would lead to agreement, we apparently do not assume that the position reflects the outside world. g. Conclusion concerning justification The arguments of the above subsections lead to the conclusion that there is no principal difference between the justification of positions that are deemed ‘factual’ and positions that are deemed ‘normative’. In both cases a person is absolutely justified in accepting such a position if this position fits in a coherent position set held by this person. Because no position set held by an actual person will be coherent in the sense defined above, a less absolute notion of justification is needed. In the spirit of the arguments above, this would be that a person is justified in accepting a position if this position is included in the set of positions actually held by this person, under the assumption that this person is not aware of required changes in his position that would make him reject this position.74 The same point can also be made differently. A person is justified in accepting a position if this position is included in his actual position set and if this position set also includes the metabelief that the set is sound with respect to the first mentioned position. See figure 1. 73 This comes near to Pierce’s circumscription of truth: ‘Truth, what can this possibly mean except it be that there is one destined upshot to enquiry with reference to the question in hand.’ Quotation taken from RL Kirkham, Theories of Truth (Cambridge: MIT Press, 1992) p 81. See also Haack, Philosophies of Logic, pp 97-8. 74 Additional refinements are possible and in the end necessary, in particular refinements concerning the necessity of making additional inquiries. For instance, a person may believe something because he did not make observations that are obviously relevant. Then this person would not be justified in accepting a position if he has is serious reasons to assume that such observations are both realistically possible and may change his beliefs. position set S Belief 2: Belief 1: … ‘S is sound with respect to belief 1’ Figure 1: Role of a soundness belief For the present purposes the most important point is that all three accounts of being justified in accepting a position do not distinguish between ‘factual’ and ‘normative’ positions. The first conclusion that follows from this insight is that the alleged ‘gap’ between ‘is’ and ‘ought’ does not a priori exclude the possibility of a normative science. The reasons mentioned under 2a, why a normative science would be impossible, turn out not to be very strong. The second conclusion that follows is that the very abstract method for science ‘Develop a coherent set of positions which includes a position concerning the question you want to answer’ is the same for all sciences, including normative ones. At first sight the prospects for a normative science, including a normative legal science, are therefore good. Still, there is a possible problem that has not been dealt with yet, namely that science is only possible when there is a possibility of agreement. Whether a normative science can, in the end, lead to agreement is not an issue that can be dealt with in an a priori fashion. Clearly, agreement and the lack thereof both occur in matters that are traditionally taken to be ‘factual’ and in matters that tend to be taken as ‘normative’. The issue of agreement will therefore be addressed in the next section, in which I will make a little more detailed proposals for normative legal science and its method. The question after the method for legal science can be answered in a very abstract way. There is only one possible reason why a position concerning the contents of the law should be accepted and that is that this position fits in a coherent theory of everything held by the person who is to accept this position. In a sense, this is the only thing which can be said about the method of legal science, or actually any science or method to arrive at justified positions, including knowledge. All other views depend on formulating part of such a coherent position set and are in a sense ‘subjective’, because any actual position set is a set held by a particular person. This subjectivity indicates the main problem that is to be overcome in the development of a normative legal science, or in fact any science. The very abstract ‘method’ for legal science may be illuminating, but because of its high level of abstraction, it is not very attractive. Therefore I will make a proposal for a more concrete method of a truly normative legal science in the following subsections, but not without warning in advance that such a proposal should be taken as part of an all-encompassing theory of everything that aims to be coherent, but will in practice always fall short of this ideal. The method for a truly normative legal science is therefore an hypothesis, comparable to other scientific hypotheses: it is falsifiable by showing that the subset of positions of which it is part is not sound with respect to the method of legal science. a. Assumptions At the beginning of this paper I claimed that the method for a branch of science depends on what one takes science to be, on the object of the science in question, on the questions that one asks about this object, of the view one takes on how answers with regard to such questions can be found, etc. The second section of this paper has put this claim in perspective: scientific method is part of a position set in which it is directly or indirectly linked to many other positions, and the mentioned ones belong to them. In section one I have formulated provisional views concerning these issues, and I will elaborate some of these views here. One assumption is that science is a collaborative enterprise aimed at gathering and systematising knowledge. It is therefore not aimed at bringing about particular results other than knowledge, although the results of science may be very useful for practical matters. From this assumption follows that science is only possible if there is a possibility of agreement on the object of the science. It also follows that scientists should to a large extent agree on what count as reasons in answering questions about the knowledge object, or- if such agreement is still lacking – about the standards by means of which views about good reasons can be evaluated. So the possibility of a science requires possible agreement in at least two ways: concerning the ultimate results of the science and concerning the methods of the science or the standards for evaluating methodological proposals. The second assumption about which I will be brief after the discussion of section 1d is that the law itself is normative. It is the answer to a normative question and in particular the question ‘Which norms should be enforced collectively?’ The desirability of enforcement distinguishes the law from morality of aspiration75 and the collective nature of enforcement distinguishes the law from (some other parts of) morality. That the state was not mentioned as the actor who takes care of the enforcement is because I do not want to limit the existence of law to situations where there is a state. b. The standard to determine what the law is To determine which norms should be enforced collectively we need standards and facts. I take it that the relevancy of the facts is determined by the standards.76 Any choice for a standard or a set of standards will be controversial. Therefore I will refrain from formulating such a choice here. It should be emphasised, however, that the adoption of integrated coherentism does not automatically lead to the Dworkinian form of coherentism that goes under the name of law as integrity.77 Integrated coherentism is well compatible with the acceptance of a plurality of independent78 values, which need to be balanced in concrete cases, and with multiple standards that govern this balancing. The ‘only’ demand is that the values and the ways in which they are balanced are coherent in the sense that these positions fit in a coherent theory of ‘everything’.79 For the purpose of this paper I will assume that the standard that would be adopted for the determination of which rules should be enforced collectively aims at the promotion of long term happiness of sentient beings. Let us call it the H-standard. 75 See Fuller, The Morality of Law, chapter 1. 76 Notice that this is an assumption about the ‘logic of justification. This ‘logic’ is as much part of the position set as the standards and the beliefs about facts, and is therefore amenable to revision too. 77 See R Dworkin, Law’s empire (London, Fontana, 1986) chapters 6 and 7. 78 With ‘independent’ I mean in this connection that the values are not ordered in the sense that some of them are merely means to realise some others. 79 A more extensive argument to this effect can be found in my Studies in Legal Logic, pp 64-67. c. Implications for legal method If the H-standard is appropriate to evaluate actions, including the adoption of a position on which norms should be enforced collectively, it is easy to determine what count as relevant facts to determine the contents of the law. Precisely those facts are relevant which concern the consequences for happiness of the collective enforcement of norms. Legal method consists therefore essentially of the methods to determine the consequences of collective behaviour for the long term happiness of sentient beings. How these consequences can best be determined is an interesting question itself, but the obvious candidates seem to be the methods employed in psychology, (evolutionary) biology, sociology, and possibly economics. Traditional legal methods that work with interpretation of authoritative texts and special forms of legal reasoning are at first sight completely irrelevant. However, there are some complications. Because the law has as one of its main tasks to coordinate human behaviour80, it is important, as Fuller has pointed out, that legal rules can easily be known, and that people will usually agree about the contents of the law. In practice this means that most law should be positive law.81 And if that is the case, at least those traditional legal methods that aim at the identification of legal rules that can easily be identified can be part of the method for a truly normative legal science. Probably this includes the reading and literal interpretation of traditional legal sources such as legislation, treaties and case law. It certainly does not include the Herculean labour proposed by Dworkin82 to make a consistent story out of apparently inconsistent legal materials. Positive law, that consists of the rules contained in the traditional legal sources, makes ‘real’ law, law in the sense of rules that should be enforced by collective means, if and to the extent that it contributes to legal certainty. If the positive law requires an interpretation that is so hard to find that it needs a judge of Herculean powers, it does not contribute to legal certainty. Therefore, such positive law does not make ‘real’ law. For the same reason legal method does not include ingenious arguments that produce results which hardly anybody would have expected. It does include, however, the setting aside under 80 I assume that such coordination produces happiness, but this assumption can and should be tested for a fullblown coherent acceptance set. 81 This is a very short abbreviation of Fuller’s argument in chapter II of The Morality of Law. That argument shows, moreover, that there are more demands than merely that the law is ‘positive’. 82 Chapter IV of Taking Rights Seriously, and chapters six and following of Law’s Empire. circumstances83 of rules of positive law if these rules literally produce unhappy results in concrete cases. Apart from legal certainty there is another reason why law consists by and large of positive law, and that is that positive law often has a democratic legitimation. Not that democracy in itself is a reason to assign positive law the status of law in the sense of collectively enforceable norms. Democracy is important if and to the extent that the adoption of positive law that was brought about in a democratic manner leads to more happiness than law which was not brought about democratically and which was only accepted because of its contents. To the extent that democracy matters, the intention of the democratic legislator may also be relevant for the interpretation of legislation. d. The possibility of agreement Any science in the sense of the collaborative pursuit of knowledge presupposes that this pursuit can lead to a result in the form of (some degree of) agreement on the investigated issues. Can the method of normative legal science lead to such agreement? In the end, the facts must decide on this question. Will legal scientist in the long run converge on which rules should be enforced collectively? There is reason to be optimistic, however. The very fact that the law must be easily recognizable and that this determines the contents of the law to a large extent, makes it plausible that legal scientist, just like ‘normal’ users of the law will agree on the contents of the law. On the assumption that the ultimate standard for what counts as law is the H-standard, it would be very hard to determine whether a particular set rules is in this respect the best if it were not plausible that some set of easily recognizable rules are probably also the rules the collective enforcement of which satisfies this H-standard. This makes that the rules that ‘happen to exist’, the positive law on a plausible interpretation of it, is a good candidate to be the law that satisfies the H-standard. This very brief argument why the positive law is a good candidate for the rules that should be enforced collectively is clearly in need of elaboration. That, however, is a topic for additional research. 83 The circumstances in question are by and large that the gain in happiness produced by setting aside the rules outweighs the unhappiness produced by the increased legal uncertainty that follows both from the actual setting aside of the rules and of the possibility that this will happen again in the future. In this paper I have argued that the proper method for legal science depends on what one takes to be science, the nature of the law and the kind of questions addressed in legal science. I started from three assumptions, namely that: a. science is the collaborative pursuit of knowledge, b. the law consists of those norms which ought to be enforced by collective means; c. the H(appiness)-standard provides the proper standard to determine what ought to be done. On the basis of these assumptions it was argued that: 1. legal science, in the sense of a description of the law, is not impossible for the reason that it is a normative science; 2. in abstract the method of all sciences, including legal science, is to create a coherent set of positions that encompasses ‘everything’ (and therefore also beliefs about the law); and 3. the proper method for a normative legal science consists primarily of the methods of sociology, psychology and economics, because the ultimate question to be answered is the collective enforcement of which norms satisfies the H-standard; the more traditional hermeneutic methods only play a role to the extent that they establish positive law that contributes to happiness by providing legal certainty. It may be useful to compare these assumptions and the conclusions that were derived from them to some alternatives mentioned above. Assumption a is not shared by Smits and Smith. This allows them to believe in the possibility of a legal science, while denying that there is a proper knowledge object. They differ on the kind of question to be addressed by legal science, however. According to Smith the proper question concerns the contents of the law, while according to Smits the question to be answered is what the law should be. Assumption b is widely rejected. Most authors, including Mackor and Smits, (implicitly) assume that the law exists as a matter of social fact. This usually, namely if legal science should answer the question after the contents of the law, leads to the conclusion that the proper method for legal science is hermeneutic. This view has difficulties with explaining the normative and evaluative nature of much legal reasoning, however. These difficulties disappear if one assumes with Smits that legal science answers the question what the law should be. On this latter view, the role of hermeneutical methods becomes at first sight hard to understand. Why does the understanding of texts help in the determination of the contents of law? Assumption c is only relevant if one takes legal science to be normative. Smits, who takes this view, emphasizes that there will be no agreement on the proper standard.