Raz’s appeal to law’s authority

Abstract

Joseph Raz’s Argument from Authority is one of the most famous defences of exclusive positivism in jurisprudence, the position that the existence and content of the law in a society is a wholly social fact, which can be established without the need to engage in moral analysis. According to Raz’s argument, legal systems are de facto practical authorities that, like all de facto authorities, must claim legitimate authority, which itself entails that they must be capable of being an authority. Further, once we properly understand what constitutes practical authority, as captured by Raz’s service conception, we realise that the directives of any authority (including the law) must be wholly identifiable without recourse to moral analysis. While the argument has previously been criticised on the grounds that the law does not claim legitimate authority, and further that the service conception of authority itself is inadequate, we argue here that the argument is actually in a worse position than these concerns recognise, for it relies upon the mistaken principle that a sincere belief or claim that p guarantees p’s conceptual possibility.

According to exclusive positivism, the existence and content of the law in a society is a wholly social fact which can be established without the need to engage in moral argument or analysis. Accordingly, exclusive positivism contradicts all rival theories of jurisprudence which allow (or, necessitate) that legal validity depends upon moral considerations, including inclusive positivism.¹ One of the most prominent arguments in favour of exclusive positivism is Raz’s (1985) Argument from Authority (hereafter, AA), which aims to demonstrate that a suitable understanding of the concept practical authority ensures that the content of authorities’ directives can be identified without recourse to moral analysis.²

In brief: (i) Legal systems are de facto (practical) authorities; (ii) In order to be a de facto authority, one must claim (and/or be considered) to be a legitimate authority; (iii) In order to sincerely claim (or be believed) to be a legitimate authority, one must be capable of being an authority; (iv) To be capable of being an authority, legal systems must possess certain properties, dictated by the purpose of authorities; and, (v) The purpose of authorities, captured by the service conception, requires (among other things) that the content of authorities’ directives be identifiable without resorting to moral analysis. Thus, in virtue of being a de facto authority, legal directives can be identified without recourse to moral analysis, consistent with the claims of exclusive positivism.

AA’s success is far from uncontested, with concerns in particular having been raised over the proposal that legal systems necessarily claim legitimate authority (Dworkin, 2002; Himma, 2019), the service conception of authority itself (Perry, 1989; Waluchow, 1994), and indeed whether the law is even capable of possessing those attributes required by the service conception (Mian, 2002). Yet, AA seems to be in a worse position than even these concerns recognise, for it owes its plausibility to a mistaken principle, namely that a sincere belief or claim that p guarantees p’s conceptual possibility. Thus, as it stands, AA contains a clearly false premise. In what follows, we identify the argument’s reliance on this troublesome principle, illustrate AA’s inadequacy by appealing to analogous arguments, and then briefly consider two potential (but, ultimately, unsuccessful) fixes. We begin with a more detailed explanation of the AA.

1 The argument from authority

Raz’s argument proceeds in four stages. Firstly, it’s assumed that legal systems are de facto practical authorities (Raz, 1985: 300).³ Importantly, on threat of begging the question, what is not being assumed here is that legal systems are de facto authorities according to the service conception of authority. Rather, a broader or more intuitive notion of authority must be at play, such that both exclusive positivists and their opponents can admit the starting assumption.

Secondly, it’s proposed that in order for the law to be a de facto authority, it must either itself claim to be a legitimate authority, be thought to possess legitimacy by those it has authority over, or both (Raz, 1979: 28, 1985: 300). The rationale is practical in nature: if some entity is to effectively hold authority, it must be deemed a rightful or warranted authority by enough members of the authoritative body and/or subjects. To use Raz’s own example of government:

To hold that a government is de facto government is to concede that its claim to be government de jure is acknowledged by a sufficient number of sufficiently powerful people to assure it of control over a certain area. A person has effective or de facto authority only if the people over whom he has that authority regard him as a legitimate authority. (Raz, 1979: 28)

Interestingly, most commentators concentrate on the claim that the law itself must claim legitimate authority, rather than on the possibility that subjects’ sincere belief in the law’s legitimacy suffices to ensure its de facto authority (Dworkin, 2002; Himma, 2019). This may be due to Raz (1979: 28) himself at times suggesting it is the authority’s own claims which are of primary importance in ensuring its status, and further that he goes to some length to justify the proposal that the law itself claims legitimate authority (Raz, 1985: 300-1), while offering no similar detailed defence of the claim that its subjects must believe in its legitimacy.

In reality, however, given the premise’s practical rationale, it seems much more plausible that effective authority requires subjects to deem the authority legitimate than the authority itself. After all, these subjects need to be given good reason to adhere to the authority’s directives and keep it in place. In contrast, a de facto authority need not think they have bona fides for their directives, but rather may issue them for pragmatic reasons (for instance, the directives serving them well personally). Of course, they may need to insincerely claim that their authority is legitimate in order to facilitate the subjects’ sincere beliefs in its legitimacy, but Raz’s concern is with sincere claims or beliefs of legitimacy and not instrumental insincere claims (Raz, 1985: 301). In what follows however, nothing will hang on whether we deem the putative claims of the authority, or the subjects’ beliefs about the authority as primary. Our interest will be with what follows from any kind of sincere claim or belief about the status of some entity.

Thirdly, in virtue of the law sincerely claiming to be a legitimate authority, or subjects sincerely believing the law is a legitimate authority, we are told it must be capable of possessing authority (Raz, 1985: 301). Understanding the exact content of this claim is complicated by a lack of clarity over how we should read the modality of “capable” here—whether it is physical, metaphysical or conceptual possibility that’s at stake—and, further, how we should conceive of the relevant form of possibility. In what follows, we’ll presume that what is at stake in AA is conceptual possibility, given that Raz himself relies upon a conceptual analysis of authority to justify AA’s conclusion. Further, as is quite standard, we’ll presume that a proposition is conceptually possible if and only if the falsity of the proposition is not necessitated by the content of the concepts it contains (Field, 1993).

Rather interestingly, Raz only requires that the law be capable of possessing the non-moral aspects of being a legitimate authority in order to sincerely claim its legitimate authority. That is, it’s enough that an entity φ be conceptually capable of being an authority, though not necessarily a legitimate authority, in order to sincerely claim legitimate authority or be believed to be one by its subjects. This appears somewhat arbitrary. After all, this would ensure that an essentially omnimalevolent, omnipotent and omniscient agent could sincerely claim to be a legitimate authority whilst knowing they were essentially incapable of being legitimate. In what follows, however, we’ll keep to Raz’s own version of the premise, as our main concerns will hold regardless of whether we require de facto authorities to be merely capable of being authorities, or place on authorities the stronger requirement that they must be capable of fulfilling the moral requirements of legitimacy.

Finally, it’s proposed that properly understood, all authorities fulfil a certain mediating function, called the service conception of authority (Raz, 1985: 303-5). According to this account, the directives of an authority regarding what relevant subjects ought to do in a given case should: (i) be based, among other factors, upon the reasons which already apply to the subjects and bear on the circumstances covered by the directives (known as the dependence thesis), and (ii) not merely add a further reason for an action to be performed, but rather replace the existent reasons subjects have (often known as the pre-emption thesis).⁴ Thus, the whole function of an authority is to mediate between its subjects and the reasons which already apply to them, issuing directives so that the subjects better adhere to these reasons than if they were simply to rely upon their own counsel.

Importantly, the service conception of authority ensures that an authority’s directives must be understandable without recourse to further moral considerations or debate. Otherwise, these directives cannot be said to effectively pre-empt the subject’s existent reasons in the case and, thus, effectively mediate on the matter. If recourse to moral considerations were (sometimes) required, then the relevant subjects would (sometimes) have to rely upon both the authority’s directives and their own judgement to determine the relevant course of action. It is this consequence of the service conception which putatively shows that the content of the law is always a wholly social fact, which can be established without the need to engage in moral analysis.

While, as noted, the service conception of authority has itself been challenged (Alexander, 1990; Darwall, 2010; Perry, 1989: Waluchow, 1994), our interest here is not with the accuracy of Raz’s account of authority itself.⁵ Rather, even if the service conception is the best account of authority we have, as we’ll now show this would not suffice to entail the AA’s conclusion and constitute a defence of exclusive positivism.

2 Sincerity ≠ accuracy

The AA hinges upon the inference from an authority φ being sincerely believed (or, claimed) to be a legitimate authority to, firstly, φ necessarily being capable of possessing authority, to then concluding that as the service conception delivers the correct understanding of authority, φ must be capable of possessing the attributes of authorities required by it:

• (1) φ is a de facto authority.

• (2) If φ is a de facto authority, then φ must be sincerely claimed or believed to be a legitimate authority.

• (3) If φ is sincerely claimed or believed to be a legitimate authority, then φ must be capable of possessing the properties necessary to be an authority.

• (4) Properly understood, authority requires that any candidate authority ψ possesses attributes q₁, q₂…q_n.

• (5) Thus, φ must be capable of possessing attributes q₁, q₂…q_n.

Yet, once we fully appreciate what AA’s other premises require, it becomes clear that premise (3) is false.

As we noted earlier, premise (1) does not presuppose a particular account of authority. Otherwise, the premise would simply beg the question against those theories of jurisprudence inconsistent with the law being an authority in accordance with the service conception. Thus, (1) does no more than to highlight our presumption that the law holds the role of an authority (whatever, ultimately, we determine that constitutes).

Given this, the consequent of (2) cannot reasonably be interpreted as requiring that the relevant parties claim or believe that φ is a legitimate authority in the sense of using the same exact application-conditions of authority as determined by the service conception in (4), even if the service conception turns out to be the correct theory of authority. As we noted in Sect. 1, the rationale for (2) is that (on a practical level) an authority cannot expect to hold that status unless (a significant number of) those involved in the authority-subject relationship deem the authority legitimate or warranted in that role. Yet, this in no way ensures that the parties involved have an understanding of authority which reflects our best theory of what constitutes an authority, nor indeed that the various parties share a consistent understanding of what constitutes an authority between them. Instead, premise (2) merely requires that the relevant class of individuals believe that the de facto authority falls under the concept legitimate authority. It does not require that this decision is based upon a full and accurate appreciation of what constitutes a legitimate authority. Indeed, we need not even presume that many have a fully worked out understanding of legitimate authority.

Similarly, we may have good reason to think that for a government to effectively maintain power in a state whose majority value democracy would (normally) require a significant majority of its citizens to believe that it was elected in accordance with the norms of a representative democracy. Yet, all this requires is that the relevant parties believe that the government falls under the concept of representative democracy. This does not mean that the citizens share the same understanding of representative democracy nor that they possess a detailed understanding of the concept, let alone the best available understanding of it. Decisions over what is admitted into the extension of a concept can often be based upon an incomplete or faulty understanding of the concept.

This brings us onto premise (4). Raz’s conceptual analysis of (legitimate) authority in (4) is an attempt to explicate our best understanding of that concept, not merely to report others’ conceptions of what constitutes a (legitimate) authority. After all, explication is not equivalent to aggregating the conceptions of some relevant class of individuals. One does not arrive at our best understanding of scientific method by summing together scientists’ reflective views on what constitutes that method; rather, one looks at what scientists actually do in order to determine the scientific method.⁶

So understood, premises (2) and (4) dictate a particular reading of (3) in order to guarantee AA’s validity. In particular, in order for the attributes q₁, q₂…q_n which Raz proposes constitute authority in premise (4) to be substitutable for the relevant descriptor “the properties necessary to be an authority” in the consequent of (3), and thus derive AA’s conclusion, premise (3)’s consequent must not merely require that φ be capable of fulfilling the application-conditions that the relevant parties happen to use for authority. Rather, it requires that φ must be capable of possessing the properties actually necessary to be an authority, as authority is most accurately and fully understood. After all, it is the latter accurate account of authority which Raz’s service conception is attempting to provide us with, and not simply to collect together the hodgepodge of potentially mistaken ideas over authority which constitute the relevant parties’ current understanding of authority.

Yet, of course, it is one matter to require that an object is capable of fulfilling a certain group of individuals’ own application-conditions for a concept (which, again, may be incomplete or mistaken), and another completely to require that the object is capable of possessing the attributes required by the concept as fully and properly understood. In particular, while it is plausible that:

(C1) Some individual I’s sincere claim that some φ is a P on the basis of φ fulfilling I’s (perhaps implicit) application conditions c₁, c₂, …, c_n for P, ensures φ is capable of fulfilling c₁, c₂, …, c_n.

In contrast, it is not plausible in general that:

(C2) Some individual I’s sincere claim that some φ is a P ensures that φ is capable of possessing the attributes actually necessary to be a P.

After all, individuals are capable of making a whole host of conceptual errors. Someone may believe that a spider is an insect merely on the basis that it is a small invertebrate. However, this does not entail that spiders are capable of possessing the attributes actually necessary to be an insect (given current science, they aren’t). Such counterexamples to (C2) are commonplace. On the other hand, if being a small invertebrate constituted the individual’s understanding of insect, then it’s unlikely they would (unless momentarily cognitively impaired) make the categorization errors necessary to sincerely claim that chairs, people and dogs were insects in a literal sense.

This is why (C1) has a plausibility that (C2) does not. To use a common distinction from linguistics, while possible counterexamples to (C1) are likely to be performance errors due to tiredness or incomplete information, which can be suitably accommodated through the inclusion of relevant caveats in the principle, counterexamples to (C2) are straightforwardly attributable to competence errors, where the relevant parties make mistakes on the basis of a lack of (conceptual) understanding. These competence errors can then subsequently lead to the parties making systematic misattributions of properties. Consequently, while AA’s (3) would be plausible if it were an instance of the principle (C1), as we have seen (3) is actually an instance of the implausible principle (C2).

This ensures that even if the service conception of authority were the best account of authority we possessed, we could not then reasonably infer that the law must be capable of possessing those properties required by the account merely on the basis that individuals believe or claim that it is a (legitimate) authority. Their categorization of it as such may simply be based upon a faulty understanding of authority. Given the possibility of conceptual error, sincerity itself is no assurance of accuracy.

Consequently, as long as there is a possible discrepancy between the relevant parties’ understanding of what constitutes authority, which determined their categorisation of φ as an authority, and what actually constitutes authority properly understood, as elucidated by (4), then premise (3) is false. As a result, unless the advocate of AA can provide us with some assurance that the relevant parties have a complete and accurate understanding of what the concept authority requires, the AA is unsuccessful.⁷

Two analogous cases should help further illustrate AA’s error here. The AA is simply a version of the more general argument form, AA*:

(1*) φ holds the position or title P.

(2*) In order for φ to possess P, it must be sincerely believed (or, claimed) by a relevant class of people that φ should (or, has a right to) possess P.

(3*) If it is sincerely believed (or, claimed) that φ should (or, has a right to) possess P, then φ must be capable of possessing P.

(4*) Properly understood, P requires that possessors of P have attributes q₁, q₂…q_n.

(5*) Thus, φ must be capable of possessing attributes q₁, q₂…q_n.

While arguments of the form AA* are valid, as we’ll now see there are substitution instances of AA* in which the conclusion turns out to be false. Further, the only reasonable culprit for the argument’s failure is (3*).

Case One: Imagine that a significant proportion of our scientific community considered creationism to be a viable scientific theory, with members of the community often citing that its viability was due to its truth being entailed by the contents of the Bible. Imagine further, however, that an eminent historian of the scientific method shows us convincingly that, based upon the contemporary and historical development of science, all viable scientific theories must actually be empirically falsifiable.

Now, in this situation, although biblical scholars may disagree over whether the Bible does indeed entail creationism or not, undoubtedly it is possible that the Bible does so entail the tenets of creationism. Consequently, in this case creationism is capable of fulfilling the (perhaps implicit) application-conditions being used for the concept viable scientific theory by some in our imagined society. However, would we also be required to conclude that, in the face of our subsequent greater appreciation of the concept viable scientific theory that creationism is capable of being empirically falsified? No, of course not. Whether it is or not is an open (and important) philosophical and scientific question.

Thus, here we have a substitution instance of AA* in which the conclusion is false. Given AA*’s validity, the fault must lie with one of the premises. Yet, by presupposition the substitution instances of (1*), (2*) and (4*) are true, and so the culprit must be (3*). As suspected, premise (3*) is false just when there is discrepancy between the relevant parties’ own application-conditions for a concept and the requirements of the concept fully and properly understood.

Case Two: Imagine that in our society we have a title Great Sportsperson awarded to any sportsperson deemed deserving, as judged initially through a petition and then a mandatory ballot (we take our sports very seriously). Those who currently hold the title include Serena Williams, Lindsey Vonn, Lionel Messi, and Phil “The Power” Taylor. In virtue of holding this status and the awarding process, clearly each of the awardees is deemed deserving by (a significant portion of) the community, although when pushed voters give varying accounts of what makes someone deserving of the title: winning the greatest number of significant championships, having the greatest impact on the sport’s visibility, making the sport seem beautiful, etc. Some may even say they base their evaluation on a hunch. Further imagine, however, that having awarded these sportspeople this venerable title, one of our most esteemed historians of sport shows us convincingly that our concept great sportsperson actually places several necessary conditions on applicability, including that anyone applicable must be at least 5’9” and able to run a seven-minute mile (during their competitive years). On accepting our esteemed historian’s conclusions, must we then conclude that Lionel Messi is capable of being 5’9”, or that Phil “The Power” Taylor was capable of running a seven-minute mile in his prime? No, obviously not. However sincere I was in my decision, if it was made on the basis of the number of championships each won throughout their career, I am not then required to admit that those I deemed deserving according to that (implicit) criterion are then capable of fulfilling the criteria explicated by our best understanding of great sportsperson.

Again, the substitution instance of (5*) turns out to be false, although the argument form is valid. Further, given the presuppositions of the case, the only plausible candidate for falsehood is (3*). A sincere belief that someone is a great sportsperson does not ensure they are capable of possessing all of the attributes necessary to be one.

These cases are illustrative not only because they help isolate the error made within AA*, and consequently AA itself. They further serve to highlight how, in general, an increased understanding of a concept can subsequently lead to a re-evaluation of its extension, especially when our initial decisions regarding its extension were unreflective or based upon hasty generalisations. In the wake of our better understanding of the concept viable scientific theory and increased appreciation of the (lack of) empirical consequences of creationism, it is completely acceptable for us to subsequently conclude that creationism ought not to be considered a viable scientific theory. Similarly, given our greater appreciation of what constitutes a great sportsperson, we have the option of re-evaluating our initial categorisation of individuals as Great Sportspeople.

Such re-evaluation of a concept’s extension has of course happened in the natural sciences, famously in the case of our reassessing the status of whales as mammals, and in the social sciences with the recategorization of autism as a disorder in its own right (Rosen et al. 2021). Indeed, it is a fundamental lesson from the philosophy of language that conceptual decisions in and of themselves cannot force entities (whether social constructs or not) to have particular properties (Williamson, 2007: Ch. 3). The determination that some concept c possesses application conditions c₁, c₂, …, c_n does not in itself ensure that all those entities which were previously considered to be within the extension of c in fact fulfil the relevant conditions c₁, c₂, …, c_n, or even that they can.

It is worth noting that if Raz’s own service conception of authority were ultimately deemed correct, then arguably the re-evaluation of at least one famous putative authority would subsequently be required, thereby highlighting that authority itself may serve as a counterexample to AA*. For practicing Christians, the Bible is a de facto authority. Further, for some Christians it is deemed to be a legitimate authority because the Bible is the literal word of God. If Raz is correct in his analysis of authority, does this require us to admit that the Bible is thereby capable of adhering to the conditions on authorities provided by the service conception? No. Indeed, it patently does not, given that following many of the Bible’s directives, including the ten commandments, require engaging in moral reasoning: What, exactly, constitutes stealing and lying for instance? Answering such questions requires moral deliberation on our part. Thus, even in the case of authority itself, it is clear that (3*) is false when there is discrepancy between the relevant parties’ own application-conditions for a concept and the requirements of the concept fully understood.

As it stands then Raz’s AA is unsuccessful, for one of its premises is false. It relies upon the mistaken principle that if one sincerely claims or believes p, then p must be conceptually possible. Sincerity is no assurance of accuracy, whether to what is actual or possible.

3 Two possible fixes

Before we conclude, two possible fixes to AA are deserving of consideration, both inspired by remarks from Raz himself. According to the first, legal practitioners can be relied upon to have an accurate understanding of the nature of the law, including its status as an authority, thereby providing us with the reassurance required for (3) to be true. In contrast, the second emphasises that legal systems are necessarily de facto authorities, and thus cannot be subsequently excluded from the extension of authority even after a reassessment of our understanding of the concept, thereby differentiating AA from our counterexamples to AA* above. We’ll consider each reply in turn.

3.1 Legal practitioners as reliable judges

While it is too much to expect the law’s subjects to have an adequate understanding of the concept authority, we may hope that in virtue of being expert practitioners of the law, lawyers and judges over time come to recognise the true nature of the law as an authority and what this entails. In other words, legal practitioners are reliable judges of the content of the law, its status as an authority, and what constitutes this authority (Raz, 1985: 302). Thus, assuming that the service conception is correct, those experienced and reflective legal practitioners will sincerely believe and claim that the law is a legitimate authority in accordance with the service conception.

As a result, in virtue of the fact that legal practitioners believe and claim that the law has legitimate authority, and these same legal practitioners have a full understanding of authority (delivered by the service conception), we can safely conclude that the law is capable of fulfilling the requirements placed upon it by the service conception.

If successful, this reply would rescue the AA by ensuring that the application-conditions which these legal practitioners use in claiming the law as an authority are accurate, and thus in this case the truth of (3) would be guaranteed by ensuring that it was an instance of the plausible principle (C1), as well as being an instance of the typically implausible (C2). If instances of (3*) are false just when there is discrepancy between the relevant parties’ own application-conditions for a concept and the requirements of the concept fully understood, then we can attempt to rescue (3) by eliminating this difference for certain suitable individuals.

Unfortunately, the reply significantly overestimates the level of theoretical understanding we expect from practitioners, even highly skilled and experienced ones. One can be an experienced and skilled scientist without having a nuanced reflective understanding of the scientific method, and one can speak German to the highest possible standard without having any decent reflective theoretical knowledge of the language’s grammatical rules. This is why we need both descriptive linguistics and the philosophy of science in order to study the grammatical rules of natural languages and scientific methods, respectively. Similarly, one of the reasons we require jurisprudence as a field is that the nature of the law (including as a putative authority) is not transparent to its practitioners. If it were, jurisprudence would amount to crafting revealing questionaries and collating the responses of legal practitioners. Consequently, one’s status as a skilled and experienced practitioner of some social activity does not ensure one has a developed and nuanced theoretical understanding of the nature of that activity, let alone a full and accurate appreciation.

In fact, within the abstract sciences we even have direct counterexamples to the claim that experts have a full and accurate understanding of the concepts in the field. Dialetheists, for instance, claim that contradictions can be true (Priest, 2006), while some classical logicians claim in response this is a conceptual mistake, that by conceptual necessity contradictions cannot be true (Slater, 1995). Either the dialetheist or the classical logician must be conceptually mistaken here, but we have no reason to doubt the sincerity of their claims. Thus, even when it comes to experts, we have reasons to doubt the principle (3) relies upon: that if one sincerely claims or believes p, then p must be conceptually possible.

3.2 The necessary authority of the law

The second reply, in contrast, reminds us of the law’s status as a paradigm case of authority, such that necessarily every legal system is an authority (Raz, 1985: 300). Thus, no adequate account of authority could viably recategorize the law as a non-authority. Given that our counterexamples to AA* above rely upon the recategorization of the relevant subject φ under an improved appreciation of the relevant concept, by guaranteeing that the law could not be recategorized in this fashion as a non-authority, we ensure the truth of the conclusion (assuming the accuracy of the service conception). Thus, however our best and full understanding of authority deviates from that of legal practitioners or the law’s subjects, it can never successfully lead to the recategorization of the law as a non-authority. Combined with the service conception, this would putatively guarantee that the law’s status as an authority suffices for the truth of exclusive positivism.

Two points are important here. Firstly, the reply rescues the AA not by resolving the concerns we’ve raised over its general form AA* (after all, the counterexamples above still hold), but rather by modifying the form of AA itself. No longer does the argument’s potency rely upon the claims of the law (or, its practitioners), and the reliability of these claims as guides to the nature of the law as an authority. Rather, simply in virtue of the law necessarily being an authority, and through an appropriate understanding of authority, we can conclude that (necessarily) the law and its directives possess the properties outlined by the service conception. In other words, the modified AA becomes an instance of the valid modal form, AA^M:

(1^M) □A.

(2^M) □(A → B).

(3^M) □B.

In particular:

(1^M´) Necessarily, the law is an authority.

(2^M´) Necessarily, if the law is an authority, it possesses the properties required by the service conception of authority.

(3^M´) Necessarily, the law possesses the properties required by the service conception of authority.

Thus, the reply is tantamount to admitting that the AA as originally conceived is indeed inadequate; we cannot use the law’s claim to (legitimate) authority as a reliable guide to the nature of its directives. Rather, we must rely instead on the law’s necessary status as an authority and our best understanding of authority to establish the nature of its directives. What we now have is a direct argument from conceptual analysis on authority.

Secondly, however, while this alteration to the AA removes the argument’s reliance on the troublesome premise (3), contrary to appearances it still fails to ensure the truth of its conclusion, and thus the truth of exclusive positivism. After all, as has happened with other instances of AA^M, the argument can easily be used as a reductio on (1^M) with the addition of the further premise:

(4^M) ¬B.

Such reductios simply serve to show that we were mistaken in our initial claim that □A. Examples abound in the sciences, for instance in the case of the apparent discovery of the element monium (Fontani et al. 2014):

(1^R) Necessarily, monium is a chemical element.

(2^R) Necessarily, if monium is a chemical element, then it consists only of atoms with the same number of protons.

(3^R) Necessarily, monium consists only of atoms with the same number of protons.

(4^R) But monium doesn’t consist only of atoms with the same number of protons.

(5^R) Thus, monium is not necessarily a chemical element.⁸

In other words, postulating the necessity of the law’s de facto authority is not the quick fix that was hoped for. In order for AA^M to be successful, we would still need assurances that no countervailing evidence can be found which undermines (1^M´) and results in the argument actually being a reductio on the law’s putative (necessary) authority.⁹ On the exclusive positivist’s part, this would require either providing: (i) independent evidence to think that legal systems are indeed, necessarily, authorities, or (ii) assurance that in fact no legal directives do require recourse to moral analysis, which can only be achieved by actually looking at all such directives, not through conceptual analysis of authority.¹⁰ Only in one of these circumstances can Raz’s AA be said to provide evidence for exclusive positivism. Raz himself offers no support for either of these assurances. Whether the exclusive positivist in general, however, can provide us with the necessary assurances for the AA to succeed is another matter which we leave to elsewhere.

4 Conclusion

Raz’s Argument from Authority is still one of the most prominent and influential arguments we possess in favour of exclusive positivism, despite the misgivings some have raised over the conception of authority the argument appeals to, and its reliance on the putative claims of legal systems. This paper has argued that, even if the exclusive positivist is able to adequately address these prior concerns, the argument is in bad shape given its reliance upon a mistaken principle: that if one sincerely claims or believes p, then p must be conceptually possible. Further, while we’ve considered two possible fixes to the argument, suggested by Raz’s own discussion, both have been found to be ultimately wanting.