Bartosz Brożek1 Two Faces of Legal Reasoning: Rule-Based and Case-Based 1. Introduction In this chapter I would like to substantiate the thesis that legal reasoning is never purely rule-based nor case-based, as it always requires a kind of interplay between abstract rules and concrete legal decisions. In order to do so, I begin by formulating two thought experiments: one, in which a highly abstract normative order is imagined and its limitations analysed; and the other, which considers a normative order consisting of particular cases only and pinpoints its failure to deliver a solid foundation for legal decisions. I further argue that the problematic features of purely abstract and purely concrete normative systems detected in both experiments are also present in, respectively, the civil law and the common law traditions. I conclude by indicating that there is no functioning legal system without a constant “dialogue” between the abstract and the concrete. 2. Two Thought Experiments Let us begin by carrying out two thought experiments. The goal of the first is to understand what are the limitations of legal systems which consist solely of abstract rules. In order to determine those limitations it is reasonable to analyse an extreme form of such a system. Therefore, we will investigate some logical aspects of an imaginary legal system which has only one universal rule of conduct. This will enable us to clearly identify the problems that accompany the utilisation of any 1 Department for the Philosophy of Law and Legal Ethics, Jagiellonian University, Kraków; and Copernicus Center for Interdisciplinary Studies, Kraków. !1 normative system which consists of abstract rules only. The second thought experiment, in turn, will concern a different approach to constructing legal orders. We will imagine a normative world in which there are no abstract rules, only particular obligations. In this way, the limitations of the “concrete” legal thinking will be exposed. Let us consider first the following situation. A legal system - LS1 - consists of only one, highly universal norm, say: (N1) Good should be done.2 This norm can be formalised in the first-order extension of the standard deontic logic as: (1) ∀xO(GOOD(x)) where O is the deontic operator “it ought to be the case that”, and GOOD is a predicate which stands for ‘does good’. Let us further assume that LS1 is complete, i.e. it can serve as the basis for answering any legal question. Moreover, let us agree - for the sake of simplicity - that legal questions concern whether there exists an obligation of a concrete person to perform some particular action. Thus, a legal question may be presented as a set {O(ACTION(name)), ~O(ACTION(name))}, where ACTION is a predicate describing some particular action and name is a proper name of a concrete individual. The problem we face, therefore, is to determine, on the basis of (1) ∀xO(GOOD(x)) whether (2’) O(ACTION(name)) or (2’’) ~O(ACTION(name)) The only way to do this is to assume that the legal system LS1 establishes an obligation of an individual in the form O(ACTION(name)) if this obligation is 2 This example is taken from Thomas Aquinas’ view of the natural law. He believes that the highest norm of this normative order is bonum est faciendum, malum vitandum. Some of the commentators of Aquinas’ thought claim that from this general norm all our rights and duties follow deductively. Cf. G. Kalinowski, Le Problème de la vérité en morale et en droit, E. Vitte 1967. !2 derivable from (1), and otherwise establishes no such obligation (i.e., ~O(ACTION(name)) is true). The problem is that O(ACTION(name)) does not follow logically from ∀xO(GOOD(x)) alone, unless ACTION(x) is equivalent to GOOD(x). What we additionally need is what may be called ‘concretisation rules’. In the standard deontic logic the following rule of inference is valid:3 If ⊢ A→B then ⊢ OA → OB where A and B stand for certain courses of action. Therefore, in order to get (2’) from (1) it is necessary to establish that (3) ∀x(GOOD(x) → ACTION(x)) If doing good involves performing ACTION, then if one ought to do good, one ought to perform ACTION: (4) ∀x(O(GOOD(x)) → O(ACTION(x))) from which by universal instantiation it follows that: (5) ∀x(GOOD(name) → ACTION(name)) Since from (1), again by universal instantiation, it follows that (6) O(GOOD(name)) we may conclude from (5) and (6) by modus ponens that (7) ACTTION(name) Moreover, if there was no such deductive passage from (1) to (7), i.e. if there was no concretisation rule in the form of (3), we would be forced to conclude that ~ACTION(name). This shows that in the case of highly abstract normative system the key role is played by concretisation rules. The problem is, where do they come from? What are the bases for formulating them? Let us observe that concretisation rules such as (3) are non-normative, in the sense that they involve no deontic operators (although they are formulated with the use of normatively-loaded concepts such as ‘good’). For example, a particular concretisation rule may look as follows: (8) ∀x(GOOD(x) → PAY_TAX(x)) 3 Cf. McNamara, Paul, "Deontic Logic", The Stanford Encyclopedia of Philosophy (Winter 2014 Edition), Edward N. Zalta (ed.), URL = <http://plato.stanford.edu/archives/win2014/entries/logicdeontic/>. !3 which says that if someone does good, then she pays taxes. As soon as we consider this example, it becomes clear that the formulation of the concretisation rules is not an easy task. It amounts to imagining a deontically perfect world, i.e. a world in which everyone behaves in the desired way. It is a world in which everyone pays taxes, does not steal or kill anyone, etc. The problem is that one should rather speak of a set of deontically perfect worlds, not a unique such world. For example, in a dentically perfect world the natural environment is protected - but there are many particular ways in which such a protection may be implemented. Plastic bottles may be recycled or their production may be banned. Therefore, if doing good requires protecting the natural environment, both a world in which plastic bottles are recycled, and a world in which they are not manufactured at all, are deontically perfect. In this way we get two incompatible concretisation rules: (9’) ∀x(GOOD(x) → RECYCLE_PLASTIC_BOTTLES(x)) (9’’) ∀x(GOOD(x) → ~PRODUCE_PLASTIC_BOTTLES(x)) This point may be made also in a different way. In the theory of obligations a distinction is introduced between individual and group obligations4 . An individual obligation is a duty of one concrete person, while group obligations are to be realised by groups of people. Importantly, the category of a group ought may be further divided into two subsets: obligations that may be fulfilled by the actions of all the members of the group only (e.g., during the lecture students should be silent) and those which may be fulfilled by the action of some subgroup of the group under obligation (e.g., students should prepare the blackboard before the lecture). It is convenient to refer to the former type as proper group obligations, and to the latter as improper group obligations. Now, I posit the abstract obligation to do good is an improper group ought. Various particular good actions may be undertaken by different individuals yielding the same deontically acceptable outcomes. The protection of the natural environment may be realised by the manufactures who do not produce plastic bottles, or by the consumers who utilise them. In other words, the abstract duty to do good is multiply realisable: there is always more than one way to find oneself in (one of) the deontically perfect worlds. 4 Cf. J.F. Horty, Agency and Deontic Logic, Oxford University Press, Oxford 2001. !4 The general moral from the above considerations is that a normative system of purely abstract rules is never self-sufficient: it cannot constitute the sole basis for arriving at a particular duty of a particular person. This is the case for purely logical reasons: any system of abstract rules necessarily expresses improper group obligations and hence may be realised in a number of ways. In order to better grasp this point let us contrast normative considerations with physics. A physicist’s goal is to uncover a unique set of laws which govern the actual world; meanwhile, a lawyer or a moralist imagine sets of worlds which are deontically perfect. Those worlds differ from one another, and hence represent different sets of particular obligations of concrete persons. It follows that - to a certain extent at least - any abstract normative system must be augmented by particular decisions in particular cases. The dream of developing a complete system of abstract rules of conduct, similar to an axiomatic system, can never be fulfilled. The second thought experiment is the following. Let us assume now that in a different legal system, LS2, there are no abstract norms. What we have is only a finite number of particular cases, in which an obligation of a particular agent is established. In order to capture this, we need to use first order deontic logic: (Case 1) SELLS_GOODS(a) FOREIGNER(a) RESIDENT(a) SELLS_ALCOHOL(a) O(PAY_TAX(a)) (Case 2) SELLS_GOODS(b) ~FOREIGNER(b) RESIDENT(b) SELLS_ALCOHOL(b) O(PAY_TAX(b)) (Case 3) ~SELLS_GOODS(c) ~FOREIGNER(c) !5 RESIDENT(c) ~SELLS_ALCOHOL(c) ~O(PAY_TAX(c)) (Case 4) SELLS_GOODS(d) FOREIGNER(d) ~RESIDENT(d) ~SELLS_ALCOHOL(d) ~O(PAY_TAX(d)) Thus, LS2 addresses the obligations of four different persons, a, b, c, and d. The persons a and b ought to pay tax, while c and d have no such obligation. Moreover, we know that a sells goods, including alcohol, is a foreigner and a resident in the jurisdiction governed by LS2; b also sells goods, including alcohol, is not a foreigner and resides in the LS2 jurisdiction; c does not sell any goods, is not a foreigner and is a resident, while d sells goods, but not alcohol, is a foreigner and not a resident. Let us assume now that we have another person, e, and we need to decide whether e has the obligation to pay tax. We know that e sells goods, but not alcohol, is a foreigner and a resident: (Case 5) SELLS_GOODS(e) FOREIGNER(e) RESIDENT(e) ~SELLS_ALCOHOL(e) In order to make a rational, not a random decision in the matter at hand, we need to treat e in the similar way as a, b, c, and d. The reasonable way to proceed would be to discern some pattern in the already decided cases 1-4, or - in other words - to spell out a universal and abstract rule or rules that govern those cases. The problem is that there is no way to do it in a uniform way. The decisions in the cases 1-4 are compatible with different (sets of) rules, such as: (R1) ∀x((SELLS_GOODS(x) ∧ RESIDENT(x)) → O(PAY_TAX(x))) (whoever sells goods and is a resident, ought to pay tax). !6 (R2) ∀x(SELLS_ALCOHOL(x) → O(PAY_TAX(x)) (whoever sells alcohol, ought to pay tax). (R3) ∀x((SELLS_GOODS(x) ∧ x≠d) → O(PAY_TAX(x))) (whoever sells goods and is not the person d - who apparently enjoys a personal exemption from paying taxes - ought to pay tax). This shows that the ‘hidden rules’ which govern the decisions in LS2 may be reconstructed in various, incompatible ways; the crucial point is that all the rules (R1) - (R3) yield the same normative outcomes in cases 1-4: a and b ought to pay taxes, while c and d do not. The problem is with the new Case 5. If we reconstruct the norm governing LS2 as (R1) or (R3), we will conclude that e ought to pay tax; if instead we accept (R2) as the correct reconstruction, e will have no such obligation. This analysis may be generalised along the lines of Quine’s underdetermination thesis. Quine’s claim pertains to theoretical discourse and states that no amount of evidence (i.e., the sentences that express the facts we observe) gives rise to a unique theory. Any theory „is underdetermined by past evidence; a future observation can conflict with it. Naturally it is underdetermined by past and future evidence combined, since some observable event that conflicts with it can happen to go unobserved. Moreover many people will agree, far beyond all this, that physical theory is underdetermined even by all possible observations.“5 In the same way, no number of past and future individual legal decisions, as well as all possible decisions, determines a unique set of rules of behaviour. There is no legal system without abstract and universal rules. The two above described thought experiments give rise to the following conclusion: legal thinking cannot dispense with neither abstract rules nor individual cases. It is a constant interplay between the abstract and the concrete. Importantly, the reasons behind this fact are purely logical. Abstract legal rules do not fully determine all particular obligations, while any finite set of particular cases is compatible with many different systems of abstract rules. 5 W.V.O. Quine, “On the Reasons for Indeterminacy of Translation”, The Journal of Philosophy 67(6), 1970, p. 178-179. !7 3. Rule-based legal reasoning Rule-based reasoning is characteristic of the approach to the law embraced by civil law systems. A good example of this way of thinking is encapsulated in Robert Alexy’s philosophy of law. Alexy claims that among legal norms, which constitute any legal system, one should distinguish between rules and principles. Rules “are norms which are always either fulfilled or not. If a rule validly applies, then the requirement is to do exactly what it says, neither more nor less”6. For example, Article 347§1 of the Polish Civil Code states that “the possessor of an immovable property shall be entitled to claim the suspension of the construction of a building if such construction might infringe his possession or threaten it with a damage.” Principles, on the other hand, “are norms which require that something be realised to the greatest extent possible given the legal and factual possibilities. [They] are optimisation requirements, characterised by the fact that they can be satisfied to varying degrees, and that the appropriate degree of satisfaction depends not only on what is factually possible but also on what is legally possible”7 . An example can be found in the Constitution of the Republic of Poland, which in Article 5 states that “the Republic of Poland shall (…) ensure the protection of the natural environment pursuant to the principles of sustainable development”. Let us observe that - at least when taken at their face value - legal rules differ from principles in character. The former are norms which ascribe rights and obligations to individuals - it is an individual who, as a possessor, is entitled to claim the suspension of the construction of a building if such construction might infringe his possession or threaten it with a damage; it is an individual, who - according to Article 415 of the Polish Civil Code - is obliged to remedy a damage caused by her fault to another person. Principles, on the other hand, express improper group obligations. When the Polish Constitution speaks of the protection of the natural environment pursuant to the principles of sustainable development, no concrete 6 R. Alexy, A Theory of Constitutional Rights, translated by J. Rivers, Oxford University Press, Oxford 2002, p. 48. 7 Ibidem, p. 47. !8 obligation is placed on an individual; rather, it is an expression of a conviction that a certain state of affairs (protection of the environment) is deontically desirable. The necessity of introducing legal principles into a legal system springs from the nature of rule-based reasoning described in the previous section. Since there is no unique deontically perfect world, but a set thereof, at least some of the legal norms must express improper group oughts, i.e. be legal principles. Because of their character, in concrete cases principles may be in conflict with one another or with a legal rule. Let us recall a famous example, often analysed in legal theory: (Vehicle in the park) A local ordinance includes a norm that bans all vehicles from entering a public park. An ambulance carrying a seriously injured person has to go to the hospital. The shortest way to the hospital is through the park. The question arises of whether the ambulance can enter the park. The rule expressed in the ordinance leads to the conclusion that the ambulance cannot drive through the park. This seems unjustified, since human life is at stake. According to Alexy, our case is a good example of a conflict between a rule (“No vehicles can enter the public park”) and a principle (“Human life and health should be protected by law”). In order to resolve this conflict one must identify the principle backing the rule. Given that the rule bans vehicles from entering a green area, it is reasonable to assume that the rule is a concretisation of the principle requiring the protection of the natural environment. Thus, the conflict we are considering is ultimately a conflict between two principles: (P1) “Human life and health should be protected by law.” (P2) “Natural environment should be protected by the law.” (P1) leads to the conclusion that the ambulance can drive through the park, while the outcome of applying (P2) is opposite. In Alexy’s theory, such conflicts are decided through the Weight Formula: Wi , j = I i ⋅ Wi ⋅ Ri I j ⋅W j ⋅ R j where Wi,j stands for the concrete weight of the principle Pi relative to the principle Pj, i.e. relative to the case at hand; Ii stands for the intensity of interference of Pj with Pi; Wi stands for the abstract weight of the principle Pi, i.e. irrespective of any !9 circumstances. Finally, Ri stands for “the reliability of the empirical assumptions concerning what the measure in question means for the non-realisation of Pi and the realisation of Pj under the circumstances of the concrete case.”8 The principle that has a greater weight prevails in the concrete case over the other principle. The last sentence needs to be stressed: the way Alexy accounts for the balancing process illustrates that an exclusively rule-based reasoning is impossible. In the case of a conflict between a rule and a principle, or between two principles, the decision determined with the use of the Weight Formula is case-relative; moreover, it cannot be reconstructed as a deductive argument in any non-trivial way. The essence of balancing boils down to the determination of the intensity of interference between the two conflicting principles, as well as the reliability of the empirical assumptions one makes. This Alexian insight reaffirms our findings from the first thought experiment described in Section 2: legal reasoning based solely on abstract rules is insufficient to yield a unique answer to every legal question. The above considerations lead to one more crucial observation. The rulebased approach to legal reasoning requires utilising the so-called defeasible logic. The concept of defeasibility was introduced into legal theory by H.L.A. Hart in 1949, and has been analysed and often redefined since. For our purposes, it is reasonable to define defeasibility as a feature of rules: A rule of the form A => B is defeasible i it is possible that although A obtains, B does not follow9 . From this definition, it is clear that defeasible rules cannot be modelled with the use of the material implication of the classical logic. A different formal system is called for, and, in fact, a number of such formalisms have been developed10 . An example is a system proposed by Henry Prakken and Giovanni Sartor11. On their account, the given set of premisses, where legal norms are expressed as defeasible 8 R. Alexy, “On Balancing and Subsumption. A Structural Comparison”, Ratio Juris, vol. 16, no. 4, p. 446. 9 Cf. B. Brożek, “Law and Defeasibility”, Revus, 23 (2014), pp. 165–170. 10 Cf. H. Prakken, G. Vreeswijk, “Logics for Defeasible Argumentation”, Handbook of Philosophical Logic, vol. 4, eds. Dov M. Gabbay et al., Kluwer Academic Publishers, Dordrecht 2002. 11 Cf. H. Prakken, Logical Tools for Modelling Legal Argument. Study of Defeasible Reasoning in Law, Kluwer Academic Publishers, Dordrecht 1997. !10 implications, serves to construct arguments, which often lead to conflicting conclusions. In the case described above, one can develop an argument based on the principle that human life and health should be protected by the law, to the effect that the ambulance can enter the park, as well as an argument based on the principle that the natural environment should be protected by the law, with the outcome that the ambulance is banned from entering the park. Those two arguments attack one another, and the goal is to determine which one prevails. The conclusion of the winning argument becomes the logical conclusion of our case (of the set of premises we have). However, the determination of which of the arguments prevails is, in general case, based on an extra-logical considerations (such as the Weight Formula). The use of defeasible logic has a number of interesting features. First, it is a natural way to model conflicts between legal rules and principles. This cannot be easily done in the classical logic, since it embraces the ex contradictione quodlibet inference rule - once a contradiction is established (e.g., the sentences “The ambulance can enter the park” and “It is not the case that the ambulance can enter the park” are simultaneously derived), anything follows from the given set of premises (e.g., that taxis can enter the park, that the constitution is not binding, or that God does not exist, etc.). Meanwhile, the defeasible logic under consideration has a built-in mechanism for handling such contradictions. Second, the use of the defeasible logic makes it possible to preserve a structural resemblance between a legal norm and its formalisation12 . Let us assume that - in the Vehicle in the Park case - the legal rule “Vehicles are not allowed into the park” is valid, but on the basis of the principle which requires human life and health to be protected by the law we make an exception for the ambulance carrying a seriously injured person. If we attempted a formalisation with the use of the classical logic, we would be forced to include this exception - and any other such exception! - in the formulation of our rule (“Vehicles - with the exception of ambulances carrying seriously injured persons - are not allowed into the park”). The defeasible formalisation does not require such a manoeuvre. We may stick to the “original” formulation of the rule, 12 Cf. B. Brożek, Defeasibility of Legal Reasoning, Zakamycze, Kraków 2004, pp. 143-145. !11 while the exceptions are introduced in particular cases through balancing of competing arguments. The rule-based approach to legal reasoning, which posits that a legal system is a set of abstract norms, is compatible with the utilisation of defeasible logic. On the one hand, this formal tool - in contrast with the classical logic - makes room for decisions which are case-relative. Since - as I have argued in Section 2 - no set of abstract rules can fully determine answers to all possible legal questions, this must be reflected in the formalisation of rule-based legal reasoning. On the other hand, defeasible logic also enables the preservation of the structural resemblance between legal norms and their formal counterparts. When formalising a legal norm, there is no need to incorporate all the possible exceptions into its logical formulations. In this way, the idea that a legal system consists of explicitly introduced norms is preserved at the logical level13. 4. Case-based legal reasoning The basic tenet of the case-based approach to legal reasoning - highly characteristic of the common-law systems - is that legal decisions are made on the basis of previously decided cases (precedents). In the common-law it is called the doctrine or the rule of stare decisis. The classic Blackstone’s Commentary defines it in the following way: The doctrine of the law then is this: that precedents and rules be followed, unless flatly absurd or unjust; for though their reason be not obvious at first view, yet we owe such a deference to former times as not to suppose that they acted wholly without consideration14 . What does it mean, however, to follow a precedent in the case at hand? It is assumed that each precedent consists of two parts: ratio decidendi and obiter dicta. 13 Cf. ibidem, p. 145. 14 W. Blackstone, Commentaries on the Laws of England, vol. 1, New York 1827, p. 47-48. !12 Ratio decidendi is the foundation for the decision in the given case, i.e. whatever aspects thereof justified the ruling; obiter dicta is everything else, i.e. those features of the case that did not influence the ruling. Let us have a look at a classic case from the English law: Mrs Donoghue went to a cafe with a friend. The friend brought her a bottle of ginger beer and an ice cream. The ginger beer came in an opaque bottle so that the contents could not be seen. Mrs Donoghue poured half the contents of the bottle over her ice cream and also drank some from the bottle. After eating part of the ice cream, she then poured the remaining contents of the bottle over the ice cream and a decomposed snail emerged from the bottle. Mrs Donoghue suffered personal injury as a result. She commenced a claim against the manufacturer of the ginger beer15. Mrs. Donoghue’s claim was successful. The House of Lords decided that Stevenson is liable for the injury she suffered. What was the ratio decidendi here? The simple inspection of the facts of the case show that the decision of the House of Lords is compatible with many rules: that Stevenson (i.e., a particular entrepreneur) is liable for the injury suffered by Mrs. Donoghue (i.e., a particular consumer); that Stevenson is liable for the injury suffered by anyone in connection with a defective product he manufactured; that producers of beer are liable for the defects of their products; that manufacturers of any goods are liable for the defects of their products; etc. Of course, the House of Lords did provide a justification for their decision. As Lord Atkin put it: The rule that you are to love your neighbour becomes in law you must not injure your neighbour; and the lawyer's question "Who is my neighbour?" receives a restricted reply. You must take reasonable care to avoid acts or omissions which you can reasonably foresee would be likely to injure your neighbour. Who then in law is my neighbour ? The answer seems to be persons who are so closely and directly affected by my act that I ought reasonably to have them in contemplation as being so affected when I am directing my mind to the acts or omissions which are called in question16 . 15 http://e-lawresources.co.uk/Donoghue-v-Stevenson.php 16 Ibidem. !13 Therefore, the reason behind the decision of the House of Lords seems to be the rule that a manufacturer of goods is under a duty to take reasonable care for the consumers of his products. This means that some of the particular features of the Donoghue vs. Stevenson had no bearing on the court’s decision: the fact that the defective product was beer, that the plaintiff was a woman, or that Mrs. Donoghue’s injury was psychological rather than physical in nature. This shows that the doctrine of precedent requires a constant interplay between particular cases and abstract rules, since the sole facts of the case are not sufficient to justify the decision. To further illustrate this point it is reasonable to consider another aspect of the common-law approach to decision-making. i.e. the process of distinguishing. It boils down to declaring that the case at hand is dissimilar to some previous precedent, and hence its ratio decidendi does not have to be followed. Let us consider the following case: Mr. McTear, who smoked cigarettes produced by Imperial Tobacco, was diagnosed with lung cancer in 1992 and died the following year. His wife filed a suit against the tobacco manufacturer, seeking damages for Mr. McTear’s death. One of the questions before the court was that of Imperial Tobacco’s duty of reasonable care for their customers. Should the court follow the precedent set forth in Donoghue vs. Stevenson and hold Imperial Tobacco liable for the death of Mr. McTear?17 Prima facie, it seems that the situation in the case under consideration is analogous to the Donoghue vs. Stevenson: a manufacturer did not exhibit the adequate duty of care and delivered a product that caused the death of Mr. McTear. However, in their ruling the court noted that “there is no breach of a duty of care on the part of a manufacturer, if a consumer of the manufacturer's product is harmed by the product, but the consumer knew of the product's potential for causing harm prior to consumption of it. The individual is well enough served if he is given such information as a normally intelligent person would include in his assessment of how he wishes to conduct his life, thus putting him in the position of making an informed choice.18” In other words, the court distinguished between two kinds of situations: when the product causes harm and one cannot reasonably expect of the consumer to know it, and when the product is harmful, but the consumer has all the 17 http://swarb.co.uk/mctear-v-imperial-tobacco-ltd-ohcs-31-may-2005/ 18 Ibidem. !14 information required to be aware of this fact and make an informed decision whether to use the product or not. Once again, we can see the interplay between cases and abstract rules. McTear vs. Imperial Tobacco is a particular case which influences our understanding of the rule governing an earlier precedent. The facts of Donoghue vs. Stevenson together with the explicit statements of Lord Atkin are compatible with two different rationes decidendi: that a manufacturer is always liable for the damage caused by their products and that a manufacturer is liable for such damages only if the consumer is (reasonably) unaware of the potentially dangerous features of the product. The process of distinguishing - such as applied in McTear vs. Imperial Tobacco - serves to state more clearly rationes decidendi of precedents. As we have seen in the second thought experiment of Section 2, any number of previously decided cases is compatible with many mutually exclusive rules of conduct. The decision in Donoghue vs. Stevenson was compatible with many rules, such as: R1: “A manufacturer is always liable”; R2: “A manufacturer is liable only when a reasonable customer is not informed or cannot easily foresee the danger associated with using the product”; R3: A manufacturer is liable only when the particular customer, e.g. Mr. McTear, is not informed about the danger associated with using the product”; R4: “A manufacturer is liable unless everyone knows about the danger associated with using the product”; etc. From this perspective, distinguishing may be described as dispensing with some of those rules and retaining others. The decision in McTear vs. Imperial Tobacco excluded R1, R3, and R4, but retained R2. However, if the court decided to follow Donoghue vs. Stevenson, no rule from the above set would be excluded. If, in turn, the decision was to overrule the previous precedent, all the rules would be dispensed with and a new (set of) rules introduced. !15 5. Conclusion I hope to have illustrated above the impossibility of arriving at a legal decision by purely abstract or purely concrete reasoning. Even if the civil law systems are constructed with the idea in mind that the law is a set of abstract rules introduced by the legislator, they cannot - for logical reasons - ignore solutions to concrete cases. This is reflected in the growing literature on the important role of precedents in the civil law systems.19 The theorists of the common law, on the other hand, have long struggled to explain what aspects of the precedents have the constraining power over future decisions20. There is little doubt, however, that the power lies with an abstract pattern of conduct “hidden” in the previously decided cases. The problem is that no set of precedents generates only one such pattern - again, for purely logical reasons, there are always alternative ways of “extracting” rationes decidendi from precedents. Rule-based and case-based approaches may be two different sides of legal reasoning, but they are sides of the same coin. 19 Cf. D.N. MacCormick, R.S. Summers, Interpreting Precedents. A Comparative Study, Ashgate, Dartmouth 1997, passim. 20 Cf. J. F. Horty, “Rules and Reasons in the Theory of Precedent”, Legal Theory 17 (2011), pp. 1-33. !16