1. Introduction
Analyticity does not inspire the same confidence today in philosophy and mathematics as it did in the early and mid-twentieth century. Today there are prominent mathematicians who deny basic principles of logic such as the law of excluded middleFootnote 1 and equally prominent philosophers who likewise doubt that analytic claims can successfully underpin logical ones.Footnote 2 In the philosophy of law, however, the analytic still does a lot of foundational work for concepts such as authority, reasons, and norms. And many philosophers of law believe that the analytic nature of these claims places them not just beyond the reach of any subsequent empirical observations but is indeed their very basis.Footnote 3 If an analytic claim about the law is invalid, it is invalid because it is incoherent at the level of its logic and not because it is incongruent with some observation about the way in which the law is experienced. This, in any case, is an influential—if not the most influential—way of looking at things in the philosophy of law today, the predominant current of which is called analytic jurisprudence. The matter at heart is an epistemological one about how we can come to know things about the world and centres on a question which might be formulated in this way: What is the relationship between logic and empirical observation? But one need not undertake this kind of groundwork in order to query a subsidiary question about analytic methodology in contemporary legal philosophy, one that preoccupies the present paper and which could be put this way: What is the relationship between phenomenology and analytic method in the philosophy of law?
To make the question practicable for a single essay, I shall restrict the examination to follow to a case study, namely, the requirement of compliance in Joseph Raz’s theory of mandatory norms as it concerns reasons for action. One ‘complies’ with a reason for action when one acts for it in doing as it requires, rather than for some other reason (or for no reason). I shall explain the requirement in greater detail throughout this essay. For now, I note that it is an ideal candidate of study because, one, it is vital to the operation of the normative system that emerges from Raz’s philosophy and, two, this normative system is arguably the most influential one in contemporary legal philosophy. If it can be shown that the requirement of compliance is implausible because it is analytic, then we will have a case study in which the relationship between phenomenology and analytic method can be described with a standard of specificity otherwise neglected in the legal philosophical literature. This will be my point of emphasis throughout. Yet some readers will be aware that Raz’s requirement of compliance has an obvious cognate in Kant’s moral philosophy—the thesis, namely, that actions are good only if they are done for the right reason. It is also possible to read what follows as bearing upon moral reasoning more generally and the question of whether there should be constraints on how we construct, as a theoretical matter, normative requirements such as compliance. But for greater precision I shall restrict the focus of my arguments to Raz’s theory of mandatory norms and I shall try to be strict in this essay by not extending the implications of my arguments to other areas of philosophy, however obvious the connections might appear.Footnote 4
I shall need to begin in section two with an account of analyticity in philosophy more generally and then the philosophy of law more specifically. Section three then introduces mandatory norms before narrowing the analysis further to the requirement of compliance. Drawing on the insights of the philosophy of art and cognitive science I shall argue that compliance is beset by a range of epistemological difficulties. Section four reviews the implications of these difficulties for method and normativity in practical reason more generally. I offer a topology of normativity and thereafter examine the burdens of proof which any normative theory must satisfy. The broader objective of what follows is to offer a description of how products of conceptual analysis—such as compliance—interact with and are susceptible to revision by extra-logical considerations in a non-question-begging way.
2. Concepts of Normativity
As I say, some philosophers who work primarily at the conceptual level think that their arguments are largely impervious to empirical intervention. Some of the concepts involved in their work are normative and can be acted upon. A good example is a promise. When agent X makes a promise to φ then, other things being equal, X ought to φ because X promised to do so, that is to say, to φ with the motive of keeping the promise.Footnote 5 As far as the promise is concerned, the promisor should not φ due to some subsequent prudential reason related to the advantages of φing. If X φs because of a subsequent prudential reason or for whatever reason other than the brute fact that X promised to φ, then X has not engaged in promissory behaviour. X has done something else, and the rightness or wrongness of the new grounds of action can be judged on its own merits. It may well be the case that it would be good to violate a promise or perhaps do the thing promised because of a reason unrelated to the promise itself or the fact that the promise was made. But the point in these separate cases is only that X has not engaged in promissory behaviour.Footnote 6 According to the standard taxonomy in the philosophical literature, it is a conceptual truth that promises work like this.Footnote 7 That is to say, the conceptual analysis of a promise shows that an act falls under it—is a promissory act—only if a certain motive is present in the mind of agents and is responsible for the agents’ doing as they promised. The ‘certain motive’, of course, is the fact of the promise itself. On this view, one can also say that it is an analytic fact that to make a promise is to convey an intention to give oneself a reason for action such that acting for it constitutes promissory behaviour. This is what is meant by the analyticity of promises. From this it follows that acting for some other fact here constitutes non-promissory behaviour.
One might summarise this idea by saying that a promise has constituent features. Only these features constitute a promissory action. Actions without these are not of the promissory kind. This in short is also what it means for a sentence to be analytic: its truth is directly implied by the meaning of the words it contains. The idea of a promise is analytic in this same sense. The reader will notice that there is nothing special in this regard about the word ‘promise’. One may be tempted to say that employing definitions just is an analytic affair. Arguments from definition however lead us to suspect at least two missing premises in respect of analyticity in the domain of normativity.
The first would be a definitional decree or customary understanding that a promise is a norm which has the features just described, namely, that it requires one to do the act promised because it was promised and not for some other reason. I will call this the definitional premise and I will use the term ‘congruence’ when referring to the relationship between (a) the norm as a reason for action (the normative reason) and (b) the agent’s reason for action (the motivating reason). If the motivating reason is the same as the normative reason, the form of congruence is called compliance. If the motivating reason is different from the normative reason but the agent still does the act prescribed by the normative reason, then the form of congruence is called ‘conformity’. Extended descriptions of compliance and conformity will follow in section 3 but for now it will suffice to say that the definitional premise will specify which form of congruence a normative reason requires. A second missing premise specific to the role of analyticity in normativity would be that one ought to invoke and employ one norm and not some other in the first place in order for it to be relevant to action. These premises granted, the rest of the work is said to be analytic. That is to say, the normative implications are derivable from the logic of the concept of promise itself without reference to anything that might be external to it, such as its justificatory grounds, the reasons for which a promise was initially made, and so on. It will not come as a surprise that these premises are not—or at least should not be—easily granted, if at all.
What, then, is the analytic programme and how is it relevant for the law and its mandatory norms? It will be of use first to remember Frege on the issue of the analytic as he sought to demonstrate the analyticity of arithmetic.Footnote 8 He wrote that
[t]he problem becomes, in fact, that of finding the proof of the proposition, and of following it up right back to primitive truths. If in carrying out this process we come only on general logical laws and on definitions, then the truth is an analytic one.Footnote 9
The point at issue here is the one about the general laws of logic and the various definitions that either attempt to capture their functions or indeed generate them. The idea is much older than Frege, foreshadowed as it was by Leibniz’s thesis that all necessary truths are traceable back to logical identities through finite analysis.Footnote 10 According to Ian Hacking, there is discord in the ranks as to whether the analytic programme can work in this way and so underpin all necessity. While Leibniz insisted that necessary truths are those that can be derived from identities by pure logic through a finite number of steps, Frege argued that necessary truths, like those which comprise Euclidean geometry, are synthetic and not analytic, grounded in what Kant called intuition.Footnote 11 Whatever we might think of this kind of dissension amongst logicians, it appears that ‘the analytic programme as a whole has run into sand’—Fregean-type confidence about what we claim to know of the basic laws of logic has long been shaken by Russell’s paradox and, what is more, diverging axiomatic set theories have undermined the idea that we can make universal claims about even the foundational rules of mathematics.Footnote 12 The result has been that the old positivist slogan that ‘mathematical [or analytic] truths are true in virtue of the words that express them’ can today only be used in what Hacking calls a ‘light-hearted way’.
In the face of all this, one might remain undeterred. Legal philosophers may find encouragement in the possibility that though logicians have not met with success in grounding mathematics on firm analytic footing, it may still be the case that analyticity is worth keeping for foundational tasks in other areas of inquiry, such as those which concern the more abstract foundations of law: language, practical reason, and normativity. As Hacking puts it, ‘drilling at mathematics with Frege’s tools, we hit impermeable matter, but that, one might feel, is no reflection on the tools’ for it may well just show that ‘mathematics cannot be fully embedded in logic’. Insofar as analyticity showed this to be the case it was a worthwhile investigative idea. But the story does not end there. Hacking reports that as early as 1936 Quine was commencing his demolition of definition, synonymy, meaning, and analyticity, and that by the mid-1960s most philosophers had come to the conclusion that the Fregean drill ‘would hardly make it through a bar of soap’. Quine’s critique was an attack on two fronts.Footnote 13 First he denied the distinction between analytic claims, which are said to be true by definition or to be such that their truths are derivable from what they mean independent from matters of fact, and synthetic ones, which are grounded in empirical facts. He argued that analytic sentences such as ‘All bachelors are unmarried’ do not actually mean anything if they are not bound to something empirical because otherwise they reduce to synonymy, which like analyticity is too obscure to be explanatory by itself. To make such sentences meaningful we need to read them synthetically, meaning that we need to register their synthetic components even when they are unstated. The claim was that ‘All bachelors are unmarried’ would mean almost nothing were it not for the referent which underwrites the meanings of both ‘bachelor’ and ‘unmarried’, where ‘referent’ refers to the brute ontology of the objects which we designate with the words ‘bachelor’ and ‘unmarried’. Here the meaning of the definiendum (the word: ‘bachelor’) is inextricable from the actual referent or definiens (the object: ‘an unmarried person’). Reference becomes object-dependent and in turn loses its analyticity because it presupposes an empirical understanding of the actual object to which the word ‘bachelor’ refers. And insofar as empirical observations are never necessarily true, the kind of analyticity which ‘All bachelors are unmarried’ is meant to represent collapses. The second front which Quine advanced targeted meaning. Drawing on the work of Pierre Duhem, in what was soon called the Duhem-Quine thesis in a related debate, Quine argued that no single sentence could be verified for its truth or falsity individually.Footnote 14 The truth or falsity of a claim—any claim, logical or empirical—could only be established in conjunction with other claims and presuppositions which comprise the same epistemic set, that is to say, the same system of belief.
Taken together, Quine’s twin arguments against analyticity helped generate what eventually came to be a holistic theory of logic and meaning around which has developed since the 1950s an entire system of philosophy.Footnote 15 At the heart of post-Quinean logic sits an idea which maintains that all claims are determined in part by the epistemological nexus in which they are embedded and so the truth or falsity of a given claim likewise depends on what one might call its epistemic neighbours, the various presuppositions and related claims that underwrite its truth or falsity. Writing in the last year of his life, Wittgenstein might have summarised this new understanding thus: ‘Our knowledge forms an enormous system. And only within this system has a particular bit the value we give it.’Footnote 16
Needless to say, these extraordinary developments in philosophy and logic have not swayed every one of the analytic tradition’s active workers. In recent years especially some have attempted to show that Quine’s critique was not as devastating as what initially seemed to be the case and that we would do better to reinterpret his work as affecting only some elements of analyticity.Footnote 17 Yet it was not to be expected that Quine would settle the matter of the analytic-synthetic discussion in its entirety and so end all philosophy which implied that the distinction was vital to its work. But it is incumbent on those of us who employ the tools of analyticity at the foundations of our work to engage the very serious doubts that have been raised in respect of the tools’ competence. The alternative is to hope that the struggles over philosophical method which are occurring in other areas of the field will resolve in favour of analyticity. Of course that will not do.
But what, returning to our initial question, have philosophers of law made of this history of analyticity? A good way to start the query is again with the name of Quine, whose frequency is a reasonable barometer of sensitivity to analytic method in contemporary philosophy. It never occurs in the work of HLA Hart, the analytic tradition’s most important philosopher of law.Footnote 18 This doubtless is for obvious reasons given both the nascent stage of contemporary legal philosophy’s formation in the mid-twentieth century and the fact that Hart and his interlocutors were interested in a very different set of questions. Notwithstanding his forays into the philosophy of language, Hart’s engagements with analytic method were subsumed almost entirely into the debate about the formal relationship between law and morality.Footnote 19 And his notorious remarks about The Concept of Law being an essay in descriptive sociology,Footnote 20 which might have firmly bound the logic of law to its phenomenology, reveal no links in his mind to the severe combustions which were occurring at the foundations of analytic method right through the 1950s, the decade preceding the publication of his major book. Oddly enough, Joseph Raz would come to criticise Hart for his commitments to naturalism and empiricist epistemology and would fault these commitments for Hart’s shortcomings in relation to the problems associated with the explanation of responsibility, the nature of rights and duties, and the relations between law and morality—‘none of them’, Raz claims, ‘was solved nor their solution significantly advanced by the ideas borrowed from philosophy of language’.Footnote 21 For Raz, Hart was too close to empiricist epistemology and naturalism, and this in turn hamstrung some of his efforts in jurisprudence. Tantalisingly close in spirit we might say to the advances Quine was instigating and which were overturning other areas of philosophy, but not, pace Raz, close enough in substance to seriously benefit from them.
In his own writings, vast and edifying as they are, Raz mentions Quine just once. It occurs in his essay ‘The Relevance of Coherence’, where Quine is mischaracterised for his ‘dual rejection of empiricism (with its belief in incorrigible foundations for all justified beliefs) and the analytic-synthetic distinction’.Footnote 22 Quine of course did not reject empiricism.Footnote 23 His target was logical empiricism, also known as logical positivism (unrelated to legal positivism), and the main recipient of his critical energies was none other than Carnap. For Quine to be misread in this way however might be thought revealing. Philosophers of law more generally have undertaken their analytic projects in general jurisprudence with an unusual unfamiliarity with the foundational literature in logic and elsewhere on analytic method—not even the literature which has sought to salvage analyticity, such as Grice and Strawson’s famous essay ‘In Defense of Dogma’ or Kripke’s Naming and Necessity.Footnote 24 There are however exceptions. Timothy Endicott offers an instructive discussion of Quine (and Williamson) in Vagueness in Law in relation to the issue of bivalence in logic but not in relation to analytic method generating legal concepts.Footnote 25 Another exception is Nicos Stavropoulos, who writes that ‘[a]nalytic truths should not be that important, ever since Quine’s famous critique of the analytic-synthetic distinction’.Footnote 26 From there the discussion moves away from the relevance of analyticity’s collapse for the basic conceptual units of law, such as mandatory norms. Andrei Marmor also offers some commentary on Quine in various works, the most relevant being the acknowledgement that ‘[c]oherence and holistic theories of knowledge have received increasing philosophical attention since the publication of Quine’s “Two Dogmas of Empiricism”’ in 1951,Footnote 27 noting further that the philosophical revolution so instigated was completed with Kuhn’s work on scientific paradigms (1962) and Rawls’ epistemic commitments to reflective equilibrium in A Theory of Justice (1971).
These writers aside, the most sustained attention to the implications of Quine for the philosophy of law occurs in the work of Brian Leiter and some of the critical commentary his writings have generated.Footnote 28 Leiter has acquired notoriety for his view that much of jurisprudence as we find it today is ‘epistemically bankrupt’ because of its reliance on the ‘analyses of concepts and appeals to intuition’, particularly in its theoretical treatment of adjudication.Footnote 29 But there is much else in general jurisprudence today that is vulnerable to basal criticisms of method besides what has been covered in the writings of Leiter and his critics. Mandatory norms are an important case in point, specifically those that are deliberately made and which require conduct unconditionally or at least almost unconditionally.Footnote 30 In the sections to follow (sections 3 and 3.1) I will describe such norms as they occur in Raz’s works and shall do so by focussing specifically on the normative requirement of compliance, which bears upon the legal norms of duty and obligation rather than, for instance, rights and powers. Then I shall offer some arguments in section 3.2 to show that the requirement is implausible because it is premised on analytic method.
3. Mandatory Norms
Insofar as they apply to action, mandatory legal and moral norms can be expressed abstractly as ‘X must φ’, where X is the agent and φ the required act (or omission). The deontic operator ‘must’ can be replaced by words such as ‘ought’, ‘should’, and other cognates; and a more complete specification would also say something about the source which issues the norm and the conditions under which the norm is relevant. A norm is ‘normative’ in that it derives its force from logic, which means that it requires the action that it prescribes in a particular way because of its formal structure rather than on grounds of, say, force or prudence. In section 3.2, under a doctrine I call ‘localism’, I shall say something about the differences that sometimes arise between responding correctly to legal and moral reasons for action. For now, I suggest that to speak of the formal structure of a norm is to specify the kinds of reasons which can be used to do the thing the norm requires. The specification moreover occurs at the second-order level because it is a regulation of other reasons, namely, those which one uses to do the norm’s prescribed act.Footnote 31 An easy example again is that of a promise. It is a formal requirement of promises that to engage in promissory behaviour one must do the thing promised because it was promised and not for some extraneous reason.Footnote 32 This is just what the word ‘promise’ means. It is in this limited sense analytic. How it comes to have this meaning and why on reflection it ought to continue regulating our usage in language in this very specific way are questions which likely comprise the norm’s most important aspects. At the moment, however, it is enough to note the use of the deontic ‘must’. It expresses the idea that a norm like a promise has certain formal requirements and that if one is to do as that norm requires correctly, then one must comport oneself with those formal requirements. A mandatory norm in the law is therefore mandatory or at least aspires to be mandatory at a few separate levels. In the first and most intuitive sense a norm can be mandatory because its breach has the potential to entail adverse practical consequences. Legal sanctions follow breaches of mandatory legal norms. But the discussion here does not focus on this more directly consequential aspect of law’s normativity. What is to be examined instead is the formal substratum,Footnote 33 focussing on the claim that if a norm requires compliance, the agent who merely conforms has reasoned in error. The notion of ‘formal substratum’ refers to the properties of concepts and, in our case, it is the properties of a normative concept, like the promissory norm, that establish the requirement of compliance.
The sensitivity to formal structure also enables an amendment to the foregoing formulation of promises and promissory behaviour. I stated that to engage in promissory behaviour one must do the thing promised because it was promised and not for some other reason. This is imprecise insofar as it situates the form of the reason for action in the fact that the promise was made rather than the fact of the promise itself. The distinction is observed if we note that the mandatory character of the norm itself is a different thing from the act which made the norm relevant. Therefore the normative requirement is that one do the thing that one promised to do on the grounds that the thing promised occurs under the rubric of the idea of a promise. This is a prolix way of saying that the norm’s formal ontology (its existence) and not its genealogy (the act from which it was begotten) is what makes it mandatory in a particular way. The promise-keeping norm, in other words, precedes the act that engages it. From formal modes of analyses, we derive more epistemic requirements at the deep level of intention, querying more exactly about the reasons that were actually used by the agent to engage the norm, the emphasis here being on the word ‘actually’. Was the thing promised done by the doer for the sake of the promissory norm, or was the thing done because of another set of reasons? Was the deed, the performance, consciously done at all as a promised thing?
I remarked that a norm’s ontology (by which I mean the logic or formal properties of a norm) and not its genealogy or occasions of engagement is what makes it mandatory in a particular way. This marks the obvious distinction between the logical properties of the norm which make it mandatory in a particular way and issues related to the act of positing a norm and therefore making some act mandatory. It will be useful to separate the issue further by explicitly specifying what it means to have something be mandatory by virtue of its extra-logical properties (for example, its occasion of engagement). I shall say that a property is ‘extra-logical’ if it cannot be shown to be true or false with logic alone. Specifying what it means to have something be mandatory by virtue of its extra-logical properties will in turn allow for greater precision in marking out the work of just the logical properties of a norm in generating mandatoriness. One intuitive way to advance the query is to ask whether a norm is mandatory because it is so treated by the society in which it obtains—that is to say, because certain things are socially demanded. This is a question about the norm’s sociality. One can say that a norm is mandatory in this and that way, where ‘this and that’ refer to social practices, the breaches of which are so widely decried that the norm comes to be recognised as ‘mandatory’. But sociality is a vast, vague, and mostly unhelpful category. It can be linguistic in that it might track language usage and the effects such usage has on behaviour; it can be political in that might say something about power differentials between disparate groups and users of norms; or, amongst other things, it can be legal in that it might hinge on converging practices, legislative acts, judicial opinions, and so on. The point in all of these social domains is that the norm takes the form that it does—it becomes meaningful—as a matter of conduct. However, nothing is inherently mandatory about a norm when it is rooted in conduct. Conduct changes and so too do the things that are considered mandatory in this way. And yet one might reply that conduct takes the form that it does because of very good reasons, like the political and legal advantages of construing a norm as requiring action in a particular way. These consequentialist arguments however are just that: consequentialist. Many if not all considerations of this sort rest on accidental phenomena for, as with conduct, the consequences of any given act can change often radically between different spaces and different times. An act which was salubrious in its consequences a century ago may well be dreadful today. The norm that makes this so is therefore contingently mandatory.
Mandatory legal norms might be understood in the way just described. But since the 1970s, in what has been called the age of discovery in the philosophy of normativity,Footnote 34 some philosophers like Raz have sought to further our understanding of norms beyond the basic awareness that different norms are called mandatory in different places and at different times. Normative theory today is grounded in the ‘logical strength of reasons’ rather than their ‘phenomenological strength’.Footnote 35 This means that in addition to whatever consequentialist considerations or, indeed, whatever other kinds of consideration might bear upon the mind of the agent, a mandatory norm also has an independent strength embedded in its formal structure. In deciding to φ because a norm R requires it, X is now understood to be acting correctly only if X is attentive to the logical or deontic requirements of R in respect of X φing, as the example of the promissory norm showed. And insofar as the primacy of the logical over the phenomenological is concerned, not much has changed in the literature since Raz’s codification of reason and normativity in Practical Reason and Norms. The same basic methods (i.e., logical rather than phenomenological priority) and the same basic conceptual apparatus (e.g., content independence, first- and second-order separations, operative and auxiliary reasons, etc.) which Raz used to advance his project remain at the centre of the philosophy of normativity today.Footnote 36
In light of this consistency, it will be well to ask about the method which locates the force of normativity in logic rather than phenomenology. As I say, I shall proceed in this respect with an examination of the requirement of compliance in Raz’s theory of mandatory norms. The requirement is arguably amongst the most basic analytic units of legal normativity and yet there is disagreement about its place in Raz’s system. Some think that compliance is not in fact a relevant normative requirement and that all that is ever argued for by Raz and others in the context of mandatory norms is mere conformity. If this interpretative objection is true, then a critique of normativity that takes compliance as its case study is a non-starter. But the objection is false and disarming it is important for the following reasons. First, the demand for compliance rather than mere conformity is what gives mandatory norms their ‘logical strength’ and also what makes them work in the Razian system as a set of reasons which is distinct from the set of reasons we may already have to do the action required by the norm, reasons that are independent of the norm. And yet if the requirement of compliance rests on flawed methodology, then the theory of normativity to which compliance is essential will in turn fail in important ways. To show this will first require that I show that compliance is indeed a normative requirement in Raz’s theory. The reader who needs no convincing in this respect may skip ahead to section 3.2.
3.1 Compliance or conformity
It will be helpful to first restate that compliance and conformity are both categories of congruence. Recall that congruence refers to the type of relation that obtains between (a) the norm as a reason for action (the normative reason) and (b) the agent’s reason for action (the motivating reason). Whenever the agent acts upon the norm as a reason to do the required deed, we have the congruence of compliance; whenever the agent uses some reason other than the norm to do the deed required by the norm, we have the congruence of conformity. The examination of congruence is an ancient one which preoccupied at one point or another almost every major philosopher from Socrates onwards, their labours roughly centring on what Hegel described as the ‘moral element’ of a deed. In the Philosophy of Right, he wrote that
the motive [Beweggrund] of a deed is more precisely what we call the moral element, and this […] has two meanings—the universal which is inherent in the purpose, and the particular aspect of the intention. In recent times especially, it has become customary to enquire about the motives of actions, although the question used simply to be: ‘Is this man honest [rechtschaffen]? Does he do his duty?’ Now, we seek to look into people’s hearts, and thereby presuppose a gulf between the objective realm of actions and the inner, subjective realm of motives.Footnote 37
The requirement of compliance takes on this presupposition and hinges on the ‘inner, subjective realm of motives’. It obtains when X decides to φ because R says that X should φ and not because it would, say, be beneficial for X to φ on some other grounds. On this view, when it comes to following a legal rule proscribing homicide, the motive must be ‘because it is the law’ and not because murder is immoral. The requirement of compliance in respect of a rule can be expressed formally by interrupting what would otherwise hold under the principle of transitivity. Where J is the justification of R and where R refers to the rule that requires φ, then the transitive relation, call it T, is ((J → R) ∧ (R → φ)) → (J → φ). The arrow indicates justifying grounds such that J justifies R in the first bracket, R justifies φ in the second, and J justifies φ in the third. Compliance under Raz’s normative theory represents an interruption of this relation such that T′: ((J → R) ∧ (R → φ)) → (R → φ). The move from T to T′ enables R itself to imply φ without reference to J or for that matter any other reason to φ but R. For the purposes of practical reason, it means that even if there is a justification—perhaps even the only meaningful justification—that would allow X to φ, compliance requires that X should φ on the basis of R rather than J. T′ represents congruence as compliance between the motivation of X and the reason for action represented by R itself. The existence of R becomes the reason for the action. This is also part of the story of how Raz argues for the idea that a mandatory norm, when it is a legal rule, is itself a reason for action and not merely a statement of a reason or reasons that an agent may already have for doing the thing required by the rule—but this we need not get into here. It suffices for our purposes to note that it follows from the foregoing framework that the mandatoriness of a legal rule requires the motive to do as the law requires because the law requires it.
Conformity on the other hand is different. Under conformity there is no requirement that the mandatory norm must be the agent’s reason for action. X might φ for whatever reason or perhaps for no reason at all (as in the case of an accident) and all will still be well as far as the norm is concerned. Accidental conformity or conformity on grounds unrelated to the norm itself does not constitute a breach of the norm. An example of the foregoing is as follows. Conformity to a mandatory legal rule requiring drivers to stop at red lights would involve a situation where a driver stopped at a red light because, say, their passenger requested that they do so; here there is conformity to the law’s requirement to heed red lights but no compliance; instead, there is compliance with the passenger’s wish and mere conformity to the law’s requirement. Had the driver stopped at the red light for the reason that that was what the law as law required, there would be compliance with the law’s requirement and conformity to the passenger’s wishes. In both cases there is congruence.
Such is the distinction between compliance and conformity, and some readers will recognise that it vaguely tracks a much older conversation about command and counsel that goes back at least to Hobbes. Hobbes in Leviathan (part II, chapter 25) and then Locke in The Second Treatise (§ 87) were concerned about the ways in which authority could rightly claim to displace the judgement of an autonomous agent,Footnote 38 and Raz would extend this discussion to encompass why the displacement of judgement must also involve an exclusion of even the justificatory reasons of a mandatory norm.Footnote 39 This exclusion is essentially what compliance demands. So which does Razian theory require for mandatory norms, compliance or conformity?Footnote 40 Practical Reason and Norms was written in 1975 but the distinction between the two kinds of congruence did not really come to the fore of the difficulties until Raz’s critics prompted him to write the postscript to the second edition of the book in 1990. The analysis that occurs in the postscript oscillates between the two possibilities, suggesting that Raz may have struggled to firmly locate the requirements of mandatory legal norms under the heading of one or the other. Indeed it is unlikely that the interpretative issue can be resolved through either just the postscript or the text of Practical Reason and Norms. Sense must be made of the distinction in light of the broader normative theory that has emerged from Raz’s system and especially perhaps through the 2001 essay ‘Reasoning with Rules’ (republished in 2009 in Between Authority and Interpretation).
What is at stake with this important interpretative difficulty is vital to the link between compliance and analyticity. This is why it is necessary to settle the question of interpretation regarding the place of compliance and conformity before turning to the more serious trouble over analytic method and normativity.
The question which instigates the deliberations on congruence in the postscript to Practical Reason and Norms is clear enough: ‘Is there anything wrong with mere conformity?’ If one lends support to a friend who is in need for reasons unrelated to the fact that the friend is in need and has asked for the kind of support that one could lend, has one failed in any normative way? Whatever shortcoming may or may not be at issue, it will not centre on the fact that support was given to the friend who asked for it. The support was given, so there is mere conformity to the norm to help a friend who is in need. The question is whether the friend who lent a hand in this way failed in other regards. Raz puts it like this:
The issue goes deep into our understanding of reasons for action. Do they aim at action, so that if the action occurs all is as well as it should be? Or do they aim at one’s reasoning as well, so that they demand, as it were, to figure in one’s reasoning and/or in one’s motivation? The view that reasons for action are always reasons for compliance fits well with the idea that practical reasons are guides to action. If one is not guided by them, then one is failing to behave as one should. If reasons for action are understood as reasons for conformity, then one may still talk of reasons for action as guides for behaviour, but only in the sense that, other things being equal, it is legitimate […] for them to figure in one’s reasoning or motivation. They are guides in the sense that the Michelin guide to Paris is a guide. I may use it, but I do not have to.Footnote 41
There is a good case to make for either possibility. In favour of understanding reasons for actions as requiring compliance, one might say that a failure to comply is tantamount to a failure of sensitivity, where the insensitivity could be legal, moral, political, aesthetic, or from whichever other plane that generated the reason for action. If, for example, the reason for action is aesthetic but one does the required deed for political reasons, then there are grounds to allege bad faith.Footnote 42 Examples are easy to come by. Should one express appreciation for an aesthetic artefact like a poem because, say, it would impress one’s peers or save one’s life or somehow advance a laudable political project, then one might argue that the act’s demerit—that is to say, the demerit of the expression of appreciation for the artefact—is an aesthetic one even if, for instance, there are good grounds to claim social or political merit or even just the merits of staying alive for the same act. The actual content of the expression of appreciation is not the difficulty: whether for instance the poem reflects this or that thought or value. The issue is that the reasons used to act, that is to say, the reasons to express appreciation, are incorrect on formal rather than substantive grounds because they fail to engage by kind in an apples-and-oranges way. On this account, one may argue that compliance ensures that an agent engages the distinct modalities or means of reasoning correctly in such a way that aesthetic reasons for action are engaged aesthetically, legal reasons for action legally, and so on and so forth. Whether compliant congruence is a good normative requirement returns us to the question of bad faith, but as the example of staying alive showed, sometimes bad faith is a very good thing. The example showed that an aesthetic demerit of an act can be overridden by the same act’s merits in respect of, for example, staying alive, assuming that (a) staying alive itself is a good thing that is better than aesthetic sincerity and (b) staying alive is itself not an aesthetic act. But this model of thinking about reasons for action—and the nature of reasons more generally—may well be wrong at its foundations, a point to which I shall return in section 3.2.
To make sense of mere conformity on the other hand is a much more straightforward affair since it does not demand any epistemic congruence (as in the case of compliance) between the norm and the reasons an actor might use to do the thing that the norm requires. X could φ for any reason unrelated to the norm which requires X to φ. Raz argues that ‘reasons for action are, barring special circumstances, merely reasons to conform’ and that ‘what matters is conformity with reason’.Footnote 43 This no-one disputes. The point though is that mandatory norms fall exactly into the so-called ‘special circumstances’ which he has in mind.Footnote 44 This is for two reasons. First, if mandatory norms in Raz’s system did not require compliance, then they would fail to be more than ‘merely statements of what we have reason to do’,Footnote 45 which are reasons for conformity. The ‘merely’ is an implication that Raz wants to avoid but which cannot be avoided unless there exists a class of reasons that require compliance. The second reason is that insofar as mandatory norms include exclusionary reasons in their formal structures, they constitute ‘reasons for not being motivated in one’s action by certain (valid) considerations’,Footnote 46 and sometimes this amounts to compliance whenever the scope of exclusion is large enough to encompass possible conforming reasons. Exclusionary reasons, remember, occur at the second-order and are negative, which means that they proscribe acting on the bases of certain first-order reasons.Footnote 47 For these reasons, it appears that compliance will have to be a necessary component of Raz’s normative machinery. That at least is what Raz implies of mandatory legal norms in Practical Reason and Norms in 1975 and the postscript in 1990, specifically in respect of those legal rules which are deliberately made and which ‘require conduct unconditionally’.Footnote 48 The 2001 essay ‘Reasoning with Rules’ makes this clearer still where he argues that mandatory norms like rules necessitate a rupture in the principle of transitivity in such a way ‘that even the justification of rules does not bear on the desirability of the actions for which they are reasons’.Footnote 49 Even the justifications of a rule—to say nothing of other non-competitive and valid reasons—do not bear as grounds of action, for here the rule itself is a reason for action and not just a statement of what we may or may not have reason to do. One can summarise this by saying that compliance is necessary to what makes mandatory norms work as a distinct system of reasons for action in Razian normativity, that is to say, distinct from the set of sound reasons for action that we may already have to do the thing the mandatory norm requires, which is the set of reasons for conformity. From these observations follow a sense of scepticism about the possibility of whether a norm can be mandatory when all it requires is conformity. We may address part of this scepticism by positing an account of mandatory norms in which they require one to identify some reason to do as they require. We could then say that such norms would constitute what are called second-order positive reasons for action, namely, reasons to act for other reasons. The details of this theory need not be elaborated in this essay.
3.2 Compliance incoherent
The question before us is now clear enough. What, if anything, is wrong with the notion of compliance? To this we may add a second. What can the requirement of compliance tell us about phenomenology and analytic method in the philosophy of law? I remarked previously that the requirement may well be unsound in its epistemology. I have said that compliance is a form of congruence between the reason for action which constitutes the norm and the reason for action used by the agent to do the act the norm requires. The cognitive model of reasoning that underpins compliance is as follows. There exists a set of reasons S for X to φ such that for the purpose of X undertaking φ, X can mentally separate S from adjacent sets of reasons and act only upon S for the purpose of φing. S, in other words, is made disjoint (in the set-theoretic sense) from all other sets in X’s universe of discourse for the purpose of undertaking to φ. The epistemology that could enable this model of reasoning is not obvious and without some kind of elaboration its use is question-begging at several levels. I will examine two such levels: the logical and the cognitive. The questions that will show the way are easy enough to formulate. One, is the foregoing model of practical reason logically possible? That is to say, is it possible to have a disjoint set in a universe of discourse which has properties like the ones that comprise, say, classical rules of reasoning? And two, is the model cognitively possible? That is to say, can our brains even do the kind of work this model requires—can we think in terms of reasons which operate in isolation from all other reasons and subsequently get ourselves to act exclusively in terms demarcated by such a reason?
For ease of reference I will call the foregoing model of reasoning to which my questions are directed ‘localism’ and I will advance two lines of argument to show why it is unworkable. The first will come by way of some preliminaries from the philosophy of art and the second by way of some harder proofs drawn from cognitive neuroscience. If it can be shown that the described epistemology is faulty, then as far as compliance is concerned the rest is silence. Any normative theory that postulates compliance as a foundational concept will need to be adjusted accordingly. It will help to first set out some terms and then restate the argument.
3.2.1 Terms and argument
Whenever a reason is said to belong to just one particular domain of discourse, I shall call it a local reason. Thus we might say that a legal reason is local to the law, an aesthetic reason local to art, and so on and so forth. Local reasons will be said to have instances whenever they are sufficiently specific. What I mean by an instance being sufficiently specific will be clear from the context in which it occurs. Instances are specific reasons within a local domain (or locale), and the rules of reasoning therein constitute the modality. We may therefore speak of a specific law as being an instance of the locale of law, whose modality is legal reasoning; and a specific aesthetic judgment as being an instance of the locale of art, whose modality is aesthetic reasoning. None of this is to say that the rules of logic alter between modalities but only that some instances of reasons may be peculiar to the locale in which it obtains and immaterial to others. What I shall try to show with the preliminaries from aesthetics and the studies from cognitive neuroscience is that neither modalities nor the instances which occur within and across domains are independent of one another. Disjoint sets are impossible or at best unworkable and the implication shall be that compliance as a construct of conceptual analysis is an epistemological non-starter.
As an aside, localism and analyticity are not to be confused. ‘Localism’ is a term I shall use for the set of epistemological presuppositions which hold that a reason that is specific to a particular domain (e.g., aesthetics) and its associated mode of reasoning (e.g., aesthetic reasoning) can be independent of other reasons in other domains and also independent of other modes of reasoning in other domains. To put it differently, localism is the view that the instances of reason and the modality of reasoning which hold in one domain can be independent of other instances and modalities. I shall suggest that localism is untenable. Analyticity is the doctrine which holds that certain sentences are true in virtue of the meaning of the words they contain. Localism and analyticity are related in the sense that both presuppose the idea that one can have self-containment—either in the form of a sentence whose truth does not depend on things outside of it (analyticity) or in the form of a domain of discourse which does not depend on other domains of discourse (localism). The distinction between analyticity and localism in hand, I turn now to the philosophy of art.
3.2.2 Aesthetics and practical reason
Aesthetic failure is far less difficult to theorise than aesthetic success.Footnote 50 Consider the various failures of a poem that valorises something morally objectionable like racism. A useful example in this regard is Kipling’s 1899 piece ‘The White Man’s Burden’. The overriding idea of the poem is that the ‘white man’ has a moral obligation, from which he ‘dare not stoop to less’, to colonise non-whites, those ‘new-caught sullen peoples, Half-devil and half child’, and ‘Fill full the mouth of Famine And bid the sickness cease’. The effort we learn is thankless because once the ends are near the colonised subject’s ‘sloth and heathen Folly Bring all your hopes to nought’. Some interpreters have charitably read the poem as encouraging Victorian philanthropy and not colonialism of the straightforward racist variety and others have claimed to find irony in it but the poem is clear enough and Kipling’s other writings on colonialism make the standard reading more credible still. The point in any case is that one cannot understand this poem as an aesthetic artefact without also understanding it as a moral one. A reader who purports to understand the poem but engages it on terms restricted to its iambic trimeter or imagery, to take two so-called formal aesthetic elements, we would say has simply failed to get it. Form, as philosophers of art have long pointed out, is constitutive of content—in the case of poetry: tone, rhythm, rhyme, metre, syntax, assonance, grammar, punctuation, and other so-called formal elements must be understood, if a poem is to be understood at all, as being ‘generators of meaning, not just containers of it’ and that ‘to modify any one of them is to modify meaning itself’.Footnote 51 ‘In literary art’, as Victor Erlich puts it, ‘ideological battles are often acted out on the plane of the opposition between metaphor and metonymy, or metre and free verse’,Footnote 52 a point which Terry Eagleton, writing decades later, would argue is true of an artefact’s ‘structural aspects’ as well:
A neoclassical poem which exploits the order, symmetry and equipoise of the heroic couplet; a naturalistic drama which is forced to gesture off-stage to realities it cannot credibly bring into view; a novel which garbles its time sequence or shifts dizzyingly from one character’s viewpoint to another: all these are instances of artistic form as itself the bearer of moral or ideological meaning.Footnote 53
What can we learn from the theoretical progress concerning form and content in the arts? More specifically, what is the relation between form and content and the thesis that neither modes of reasoning, nor instances within and across each mode, are independent of one another? Erlich and Eagleton point out that it would be a failure of understanding to mark out a fundamental distinction between the formal and non-formal aspects of an aesthetic artefact like a poem. The reason for this put simply is that form is content. The aesthetic is the moral and the moral is the aesthetic in works like Kipling’s poem, where ‘aesthetic’ is typically taken to refer to ‘formal’ qualities. Aesthetic reasoning, as a mode of reasoning concerned with such qualities, cannot generate a competent understanding of an aesthetic artefact because every formal device is imbued with meaning, some of which might be moral, and would require moral reasoning, some of which might be political, and would require political reasoning, and so on. Insofar as one mode of reasoning must perforate other modalities in order to get at competence (that is to say, to understand the thing that one is reasoning about), since none exists in a vacuum, so too are the instances of one method interlocked with those of other modalities. These are the kinds of reasons that could enable a critic to argue that it would be an aesthetic failure for a poem to valorise something morally deplorable like racism.
Nevertheless, for practical purposes it is often the case that one has good reason to attend exclusively to just the aesthetic or just the morality or even just the legality of an act or artefact. These practical purposes can be myriad: in a court of law, the aesthetics of, for instance, rape are irrelevant because aesthetic valuation is not, as the trials of Oscar Wilde showed, an accepted mode of legal argumentation;Footnote 54 in moral philosophy, the lawfulness of a socially accepted practice is also often beside the point because we legally do all sorts of things as a society that we really ought not to; and in aesthetic theory, one often has reason to set aside other modalities like moral reasoning or political reasoning in order to strike at other depths—as one might, for instance, by bracketing Wagner’s nationalism in order to learn what we can from his conception of totalism in the arts,Footnote 55 or by bracketing Fragonard’s reverence for aristocracy in The Swing in order to learn what we can from its sensitivity to social narrative. But in all these cases, the bracketing of other modalities (e.g., moral reasoning) and other instances (e.g., a moral reason) into separate locales of engagement is a tactical decision that is made on a balance of reasons that centres on what is considered useful or appropriate in a disciplinary way. As far as epistemology is concerned, and in unison with the logical theories of holism one finds in Wittgenstein and Quine, the kind of localism just described holds no water. No mode of reasoning and by extension no single piece of reason is ever unadulterated by other modes and other reasons. Cross-contamination between the different rational methods is an ineradicable fact of reasoning. This is why aesthetic theorists like Andrew Britton are committed to the view that ‘the art critic is committed to totalize’,Footnote 56 meaning that one’s work must be done with a cognisance of its various epistemic and contextual contingencies. It is also why Frege argued that ‘the meaning of a word must be asked for in the context of a proposition, not in isolation’, essentially foreshadowing Britton at the formal level, and also why Wittgenstein and Quine would extend this contextualist principle to sentential and logical utterances more generally. It is too innocent, then, to think that one’s attempt at something like an aesthetic or legal judgment can be cleanly aesthetic or legal or cleanly anything.
3.2.3 Cognitive science and practical reason
There are good empirical grounds for the persistently gestaltist view of every instance or specificity of reasoning. Much of the support in recent years has come from experimental psychology and studies that investigate the neural bases of reasoning. Writing in 2008 on the psychology of moral reasoning in particular, Monica Bucciarelli and her colleagues found that ‘no simple criterion exists to tell from a proposition alone whether or not it concerns morals as opposed to some other deontic matter, such as conversation, a game, or good manners’.Footnote 57 Admitting the absence of this ‘simple criterion’ or ‘simple defining property’ enables the search for other non-atomic indicators which could, among other things, contribute to the generation of one’s reasons for action. The investigators in this particular study put their fingers on the ‘specific knowledge of your culture’, a rather vague way to say that one’s reasoning and so one’s bases for action are complicated and never unitary.
The findings from Bucciarelli also correspond nicely to an earlier negative result from brain-imaging studies by Joshua Greene and Jonathan Haidt in 2002, which showed that ‘there is no specifically moral part of the brain’ and their joint conclusion that:
morality is probably not a ‘natural kind’ in the brain. Just as the ordinary concept of memory refers to a variety of disparate cognitive processes (working memory, episodic memory, motor memory, etc.), we believe that the ordinary concept of moral judgment refers to a variety of more fine-grained and disparate processes, both ‘affective’ and ‘cognitive’.Footnote 58
This conclusion is consistent with neuroimaging work by Vinod Goel, whose fMRI studies of deduction—essentially the principle of transitivity with which we have been preoccupied—showed that reasoning with sentences that had semantic content (e.g., ‘All apples are red; all red fruit are sweet; therefore all apples are sweet’) as well as those without or, at least, with less semantic content (e.g., ‘All A are B; all B are C; therefore all A are C’) show engagement with both linguistic and spatial systems in the brain.Footnote 59 The vast spread of reasoning processes Goel describes coincides with a related argument from a further study he conducted with Jordan Grafman, which held that ‘reasoning does not stand alone as a cognitive process’ and that it is instead caught up with numerous other cognitive tasks, such as forming metaphors, developing plans, and navigating our various social and non-social circumstances.Footnote 60
These studies in cognitive neuroscience have a number of implications for the way in which we understand the actual processes of reasoning that animate practical reason and normativity. The first is that categories like moral reasoning do not operate atomically, meaning that both the mode of reasoning (e.g., moral) and the specificity of the content such reasoning produces (e.g., ‘φ is good’) are plaited with other modes and so other specificities: what, in other words, was at the crux of the prior argument, a line cast alongside Wittgenstein and Quine, against localism. What is more, moral reasoning may refer to a set of other ‘more fine-grained and disparate processes’, compounding further the problems associated with localism. ‘Moral reasoning’ here is a placeholder—what can be said of it can be said of legal reasoning as well; not because legal reasoning is a kind of moral reasoning (it is not) but instead because, and at least insofar as, both imply or rely on normative injunctions (e.g., ‘X should φ’). All of this lends support to the idea that every mode of reasoning and every specific reason is bound up with a nexus of other modalities of reasoning and specificities of reason from which it is inextricable. If this is true, we never have atomic reasons for action, atomic in the sense of a single reason or a single mode—hence the inappropriateness of localism as an expectation in any theory of practical reason.Footnote 61
If the abandonment of localism can be sustained, then the idea that one can even have compliance, as a theoretical matter, runs into difficulty. Compliance, remember, specifies that the reason or set of reasons one uses to do as the norm requires must be no other reason or set of reasons than that which constitutes the norm itself. This is nothing less than an attempt to apply localism to normativity. Compliance, moreover, requires that an agent not act upon excluded reasons which would otherwise be valid as reasons for action, even valid as reasons for the same action required by a norm. Remember also that an excluded reason refers to a reason which is eliminated from the field of valid rational motivations that an agent can rightly use to do the thing that a norm requires. Put more technically, an exclusionary reason is a negative second-order reason, which means that it is a reason not to act upon an excluded first-order reason to φ or not-φ. This involves not being motivated in one’s actions by reasons which are not explicitly stipulated by a norm when that norm is relevant. But the things one might do and the things one might say cut across a range of categories because they respond to a nexus of reasons the constituent parts of which are inextricable from one another. Every reason is bound up with other reasons: with other modes of reasoning and other instances of reason. The requirement of exclusion under compliance, however, amounts to an intransigent denial of this fact about the nature of reasons. The idea that one can isolate some reasons, exclude others, and then neatly beget a set of motivational considerations to which one can rightly resort as reasons for action is as incredible as the demand that one should view the world monochromatically when one’s visual apparatus is in fact set up to see colour. Neither formal reasoning nor our cognitive processes work like that even if, for practical purposes, we make tactical decisions to treat reasons and their concomitant sets in isolation, as we might in an art gallery or a court of law. But these are disciplinary decisions and not descriptions of the structure of practical reason.
4. Method and Normativity
It is now time to ask about method. It will be vital to retain the separation between, first, the basic justification for regarding mandatory norms as requiring one to reason about one’s actions in a particular way and, second, the way in which such norms require one to reason about one’s particular actions once the justification is in place. With this in mind, we ask: how exactly does one arrive at the idea of compliance—that is, the notion that one must do as one is required to do for the reason that one is required to do it—in explaining the distinctiveness of the reasons for action which mandatory norms are meant to occasion? Observe that the following two claims are compossible, that is to say, they can be joined with a conjunction:
(1) The requirement of a norm for one to reason in particular ways (for example, by complying) in doing as the norm prescribes can be justified on utilitarian grounds.
(2) That requirement—the mental act of reasoning about the norm and the action it requires—can operate analytically, as in the case of compliance or promising.
Observe further that it is not possible to run the following two claims together:
(3) The justification of a norm can be analytic, by which it is meant that if one is to understand and act upon the norm correctly, one must understand the norm analytically (again, as in the case of compliance or promising).
(4) The reasoning in respect of that norm can operate on the basis of utility.
As mentioned, a claim is extra-logical if it cannot be substantiated with logic alone. The thesis, found in Raz, that we ought to comply with the law rather than conform to it because it maximises obedience or utility, for instance, is an empirical claim that requires more than the instruments of logic to prove or disprove it. It is in other words ‘extra-logical’. Claim (4) is in this respect extra-logical. And if claim (3) is true and that the correct way to reason about a norm is analytically, then it is irrelevant as far as reason is concerned whether it is useful for such a norm to be viewed analytically. To reuse the example of the promise, the tension between claims (3) and (4) centres on the compliance requirement of promises on the one hand and the definitional violation of the concept on the other were one to do the intended action of the promise for some reason other than the one represented by the plain fact of the promise itself. It is possible to put this differently by saying that the bifurcations at stake—between claims (1) and (2) and between claims (3) and (4) are about segregating the justification of the norm, a matter which turns on ‘why’, from its modality, which is about ‘how’. The segregation, that is to say, between (i) why one should accept that something is a mandatory norm and (ii) how that thing operates in one’s practical reasoning once it is regarded as being justified. And yet it seems that both halves of these separations—call them the justificatory and modal branches of normativity, respectively—turn on method. If, for instance, the justification of a norm is utilitarian, i.e., to maximise economy or advance human happiness, then one must query whether it is worthy of belief, that is to say, one must scrutinise the method of arriving at that conclusion. If the justification for regarding a norm as mandatory in a particular way is empirical, then the debate will turn on the data. Without the right data, the moderate way to put the ensuing objection would be to point out that the claims being advanced are at best as good as their speculative inversions. On the other hand, if the justificatory reasons are analytic, then we are in the peculiar (or perhaps impossible) position of having tried to advance a utilitarian argument, which is necessarily contingent on extra-logical facts, through analytics.
The justificatory and modal divisions of normativity are often neglected in Raz’s theory of mandatory norms and this has had knock-on effects for methodological clarity both in Raz’s own works and those who use his conceptual apparatus throughout their own legal philosophical projects. Part of the reason for this neglect is that much is made of a similar but different distinction between the prescriptive and evaluative grounds of a norm. The difference, that is to say, between the question that asks whether a rule is binding and the one that asks whether a rule is justified.Footnote 62 The space between the prescriptive and evaluative—the so-called ‘normative gap’ and associated ‘opaqueness of rules’—is meant to describe the space between a rule’s value and its prescriptive force.Footnote 63
4.1 Topology of normativity
The distinction captured by the normative gap, as important as it has been for legal philosophy as a whole, is not the relevant one for the current examination.Footnote 64 As I say, the focus of evaluation is (A) ‘Why is X’s φing for R good?’, while the focus of prescription is (B) ‘Why should X φ for R?’. The distinction between evaluation and prescription is the classical focus of legal philosophy. The former concerns value or goodness (‘φ is valuable because it promotes communal wellbeing’), whereas the latter indicates instructions for action (‘the law requires that X must φ’). However, the distinction with which I am concerned—namely, between justification and modality—is both different and also serves other critical ends. Under justification and modality, the respective questions are (C) ‘Why is R, the norm that requires X to φ, good?’ and (D) ‘How does R regulate X’s reasoning in relation to φ?’
Table—Topology of normativity

Notice that (C) is sometimes reducible to (A) for the reason that R being good is often derivable from the good of X φing for R. This reduction of justification (C) to evaluation (A) is no surprise. For a norm to be justified the action it requires should also be worthwhile on evaluative grounds. In this case, however, the deontic ‘should’ in the preceding sentence is not ‘robust’ since it is possible for an action to have no evaluative merits on its own at all and yet be made worthwhile by virtue of the norm which requires it, in which case justification (C) cannot be derived from evaluation (A) in a non-question-begging way. Examples abound, contractual obligations perhaps being the most intuitive: contracts often require actions which would otherwise bear no value whatsoever; such actions are made good by the fact that they are required by contract. For present purposes, the more interesting division in the topology of normativity in the above table is modality (D), which is not reducible to or derivable from its seeming cognate under prescription (B). The reason for this is that it is one thing for something to be a prescriptive reason like ‘X ought to φ’ and another thing altogether to stipulate the way by which X ought to reason their way to φ through R’s formal structure (for example, by way of compliance). Nor is modality (D) moreover derivable from justification (C), for the claims that fall under the latter do just that: they justify but do not stipulate the means, that is to say, with a certain motive present in one’s practical reasoning, by which the norm is meant to operate or bear upon one’s reasons for actions. It is one thing for a norm to require that ‘X must φ’, which is a prima facie prescriptive reason (B), and another thing entirely for an additional claim to stipulate that ‘X must φ by means of compliance’, which falls under the rubric of modality (D).
Under modality, we must bear in mind what has already been said about the analyticity of compliance throughout section 3 and also, more generally, the combustions analyticity has suffered in post-Quinean logic. And here the difficulty is clear enough. If the analytic project of Razian legal philosophy—that is to say, much of contemporary legal philosophy insofar as it concerns mandatory norms—stipulates that X must φ by way of compliance, then this amounts to a requirement that modality in turn work by way of analyticity, a requirement which I have argued is like the demand that one view the world in black and white when one’s visual apparatus is set up to see colour. And yet if uniquely compliant motivation is jettisoned as a normative requirement because it relies on implausible models of reasoning and human cognition, philosophers of law will need to grapple once more with the so-called puzzle of the opaqueness of rules, the puzzle, that is to say, which derives from the claim that ‘rules, some rules, are themselves reasons, and not merely statements of what we have reason to do’.Footnote 65 For if compliant congruence is eliminated as a modal alternative to normative behaviour like rule-following, then a norm like a rule cannot be a reason for action itself and not just a statement of extant reasons since it will always rely on grounds that are external to itself—grounds which of course an agent may well employ to do the action required by the norm at hand irrespective of the fact that the norm requires it. In such a case, the norm cannot but be a statement of extraneous reasons which one might already have or at least be able to acquire independently of the norm. The trouble with the puzzle about the opaqueness of rules, at any rate, is just one further implication of the breakdown of compliant congruence as a normative requirement in an explanation of mandatoriness. But it is beyond the task here to offer an alternative solution to the disarray this inspires for legal normativity.Footnote 66
4.2 Burdens of proof in normative theory
It will be well to demonstrate as a final move how Raz might navigate the topology of normativity described in the preceding section. Remember that Raz is of the view that mandatory norms must be regarded through the prism of exclusionary reasons because otherwise they would not serve their purposes as mandatory norms.Footnote 67 Such purposes in turn are grounded in various justifications of mandatory norms, two of which Raz takes from Mill’s A System of Logic and a third which he derives therefrom as well:
By a wise practitioner, therefore, rules of conduct will only be considered as provisional. Being made for the most numerous cases, or for those of most ordinary occurrence, they point out the manner in which it will be least perilous to act, where time or means do not exist for analysing the actual circumstances of the case, or where we cannot trust our judgment in estimating them.Footnote 68
Two justifications then from Mill: first, to save time, and second, to reduce risk of error; and then a third which Raz appends, namely, to save on labour. These are not the only possible justifications available, there are others and Raz of course notes as much. It is moreover not important to the discussion here whether these are good justifications for something as legally and morally important as a mandatory norm—though as an aside one may note that they very well might not be the kinds of reason one would want regulating mandatory conduct. What matters instead is that they are consequentialistic, and the point to bear with utilitarian justifications of what is mandatory is that some if not all of them are open to empirical testing. This prompts a torrent of enquiries about method. If the claim is that rules so conceived—that is to say, conceived as constructs of conceptual analysis that require that an agent’s action be regulated by reason in a particular way, say by means of compliance—will save on time and labour and also reduce risk of error, then it is a straightforward task to test what are clearly empirical claims about actual human reasoning and behaviour. We can test to see whether a norm whose justification is offered as, for instance, an error-reducing device actually serves the purpose of error reduction when its modality is governed by compliance. What is more, the tests and the resultant evidence must stand in comparison to what might have happened with different modalities. The weight of the argument for constructing norms with some modality, like compliance, must accordingly be regarded as relative to arguments for different modalities. The reason for this is that a norm may well reduce error when it is constructed as one that requires compliance and which works by way of exclusionary reasons but it will remain an open question what the point of reference is and how mandatory norms so conceived compare against other formulations that entail different modalities. The point in any case is about the weight of the evidence before us. As I say, if no evidence is advanced to buttress empirical claims of this kind, then the claim is at best as good as its inverse. In respect of mandatory norms, pace Raz, no such evidence is on offer.
It is instructive that there is no empirical evidence for the empirical justifications that underpin the processes of modality in Razian normativity. In Raz’s theory of normativity, we are required to reason deontologically about norms (by way of compliance, for example) because it is better to do so as a matter of consequences. We are told in other words that to reason in this way about norms results in better outcomes. Yet no evidence has ever been offered that to reason deontologically in fact results in the good consequences that are claimed for it or that it produces better outcomes than reasoning about norms non-deontologically. And yet the story does not end there. One might draw in one’s horns and argue that evidence need not be empirical. And one might think that this may especially be the case if the matter at hand centres on the logical strength of norms rather than their phenomenological strength.Footnote 69 Two remarks follow.
First, we must leave untested the grounding assumption that logic need not answer to observation as otherwise even the most formal systems will require some empirical premises. Even if this assumption were false, its implications for legal philosophy and practice would be uninspiring. Few philosophers of law and even fewer lawyers would dispute the suggestion that legal concepts at least need empirical premises.Footnote 70 So the assumption about the structural separation between logic and observation we need not touch.
Second, and more to the point, it is to be emphasised that the burden of empirical proof obtains for each of the categories of normativity indicated in the Table: (A) Evaluation, (B) Prescription, (C) Justification, as well as (D) Modality. What is of special interest in the current context is modality but it will be well to quickly state why the preceding three categories require not merely foundational empirical premises but ongoing empirical examination. Consider again the basic mandatory norm R which states that X must φ by way of compliance. R constitutes the norm’s prescription and compliance its modality, its method. Asked about its justificatory grounds, the Razian theorist might cite Mill and say that R is justified because it reduces error. Asked about its evaluative grounds, they may say that it is good to reduce error because, for instance, it maximises utility or human happiness, assuming thereafter that no further regress is desirable or possible. And so when the objection is made that evidence need not be empirical, the response—now that there is a workable topology of normativity—is to ask about the normative category at issue. Is it that the evaluative claim cannot or need not be empirically examined? This would be strange. If φ is said to ultimately maximise some kind of good, like utility, then surely we can examine whether this is so at least in the context of the law and legal policy. Or is it that the justificatory grounds are impermeable to empirics? This too would be strange. If a rule is justified on the ground that it reduces error, then we can test whether it actually reduces error. So too I have said with modality. If a modality (method) is chosen because it is claimed that the rule in question would otherwise not serve its purpose (found under its justification),Footnote 71 then three questions follow. One, is the modality so chosen fit for the purposes set out by the rule’s justification, that is to say, does it in fact do something like reduce error? If the answer is yes, then the matter is still, as described, a comparative one. Two, is the modality formally possible? I argued in section 3.2.2 that compliance at least is not. And three, is the modality cognitively possible? Again in respect of compliance I showed in section 3.2.3 that there is good scientific evidence to show that it is not. Indeed the burdens of proof are manifold—they trace the topology of normativity drafted in the Table and it is submitted that they remain unsatisfied in Razian legal theory.
5. Conclusion
This paper has sought to contribute to the conversation on method in the philosophy of law by way of a case study, that of the requirement of compliance in an influential theory of mandatory norms. Elsewhere in philosophy, perhaps especially in logic and the foundations of mathematics, writers more sensitive to the history of analyticity are cautious in respect of the lingering enthusiasm for its explanatory power. In 1936, AJ Ayer could proclaim that
there is nothing mysterious about the apodictic certainty of logic and mathematics. Our knowledge that no observation can ever confute the proposition ‘7 + 5 = 12’ depends simply on the fact that the symbolic expression ‘7 + 5’ is synonymous with ‘12’, just as our knowledge that every oculist is an eye doctor depends on the fact that the symbol ‘eye-doctor’ is synonymous with ‘oculist’. And the same explanation holds good for every other a priori truth.Footnote 72
Even in 1975 Ian Hacking reported that ‘[n]o one would dare say such a thing any more’.Footnote 73 But some forty years later in the philosophy of law, the predominant current of theory is even called analytical jurisprudence.Footnote 74 The old maxim that ‘necessary truths are true in virtue of the meanings of the words used to express them’Footnote 75 endures too. In directing sceptical attention to compliance I have tried to show that at least insofar as that analytic requirement is concerned, we have good formal and empirical grounds to refrain. The immediate implications of course are limited. Compliance will become unworkable. The theory of mandatory norms to which it belongs may need to be modified also. Beyond that, however, the more general cautionary remarks about analyticity are by now famous.
