The Service Conception: Just One Simple Question

It is crystal clear that the Service Conception includes at least three conditions, what I shall call: the ‘normal justification condition’, the ‘independence condition’ and the ‘dependence condition’. The overarching rationale of these conditions is that they ensure that authority is only justified when it provides the best means for the subject to conform to the reasons for action that she actually has. However, it is difficult to clarify whether Raz implicitly presupposes a fourth necessary condition. This condition might be called a ‘reliable belief condition’, that is, that the putative subject must reliably believe that the putative authority-agent satisfies the Service Conception (or more precisely, its other three conditions). In sum, the purpose of this paper is to pose Joseph Raz one simple question: is it a necessary condition of your Service Conception, that the subject believes that the authority-agent satisfies the Service Conception? As a matter of interpretation, different parts of Raz’s work appear to lead in entirely opposite directions: some parts clearly support the reliable belief condition, others do not. Regardless of Raz’s ultimate answer, however, the question reveals a broader inconsistency. Only if the Service Conception does include the belief condition will it support Raz’s claim that authority is consistent with one’s rational ‘self-reliance’, that is, acting upon one’s own judgement (including, as to who has authority). Only if the Service Conception does not include the belief condition will it support Raz’s perfectionist account of government. It seems Raz must choose between one or other.

tion', the 'independence condition' and the 'dependence condition'. The overarching rationale of these conditions is that they ensure that authority is only justified when it provides the best means for the subject to conform to the reasons for action that she actually has.
However, it is difficult to clarify whether Raz implicitly presupposes a fourth necessary condition. This condition might be called a 'reliable belief condition', that is, that the putative subject must reliably believe that the putative authority-agent satisfies the Service Conception (or more precisely, its other three conditions). 2 In sum, the purpose of this paper is to pose Joseph Raz one simple question: is it a necessary condition of your Service Conception, that the subject believes that the authority-agent satisfies the Service Conception?
As a matter of interpretation, different parts of Raz's work appear to lead in entirely opposite directions: some parts clearly support the reliable belief condition, others do not.
Regardless of Raz's ultimate answer, however, the question reveals a broader inconsistency. Only if the Service Conception does include the belief condition will it support Raz's claim that authority is consistent with one's rational 'self-reliance', that is, acting upon one's own judgement (including, as to who has authority). Only if the Service Conception does not include the belief condition will it support Raz's perfectionist account of government. It seems Raz must choose between one or other.
In the first two sections of this paper, I outline the Service Conception's less controversial elements, and frame our central question. I then outline, with exegetical support, two very different interpretations of Raz's work with respect to answering this question. I then investigate the prospect of reconciliation, but only to dismiss it. I conclude by drawing out the implications of this inconsistency for Raz's work.

I. THE SERVICE CONCEPTION
Let us first articulate the conception of authority that is the object of the Service Conception; followed by the account of legitimacy that is the substance of the theory.
Raz starts 'with the common thought … that authority is the right to rule'. 3 'Commands' are products of its exercise that recipients, qua 'subjects' have a 'duty' to obey. 4 We should emphasise four key elements.
First, such authority is practical rather than theoretical (or 'epistemic'). It is concerned with an agent providing to a subject, through commands, reasons to act; rather than reasons to believe.
Secondly, it is de jure rather than de facto. It is concerned with an agent who has a right to rule in a manner that their command, by itself, creates a reason to so act (qua duty). This reason to act, therefore, is not dependent upon extrinsic factors such as the threat of punishment, (although such factors may create additional reasons to so act).
Third, 'duties' are a special type of reason for action, conformity with which is 'nonoptional'. According to Raz, this is because they are, in fact, the systematic combination of two reasons. First, a reason to act as directed; and secondly, an 'exclusionary' reason, not to act on the basis of otherwise valid reasons to act to the contrary. 5 In this way, the reason created by the command is not added and weighed against other reasons, but simply overrides them. Raz calls such reasons, 'protected'. 6 Finally, Raz does not expand on the conception of 'right' in play with respect to authority beyond being the 'abilit[y] to impose or revoke duties or to change their conditions of application'. 7 Definitionally, therefore, Raz takes the concept of a duty to be most primitive.
Raz's conception of authority, therefore, is the de jure ability to impose (change, revoke etc.) upon a subject protected reasons for action. The aim of Raz's Service Conception, as a theory of legitimacy is, putting aside deviant cases, 8 to establish the necessary and suffi-cient conditions for such an ability. It clearly includes the following three conditions. First, that an agent would conform better to the reasons that apply to her regardless of the command of the authority-agent, if she intends to be guided by the authority-agent's command rather than if she does not. This is dubbed, 'the normal justification thesis'.
Secondly, that the reasons that apply to the agent regardless of the authority-agent's command are such that it is better to conform to them rather than to decide for oneself, without authority. This is dubbed the 'independence condition.' The independence condition's chief purpose is to recognise that one important reason we often have for action is the promotion of our own 'autonomy'. Persons with 'autonomy' are 'part creators of their own moral world.' They not merely decide how to act for themselves but have a capacity to (partly) define, through these decisions, what is valuable for themselves by creating personal projects, commitments, causes and relationships. 9 Sometimes, therefore, the value of autonomy creates a reason to decide for oneself, even if one would otherwise better to conform to reason by following the command of another.
Thirdly, authoritative commands must derive from the authorityagent's own attempt to assess and weigh the relevant reasons that already apply to the subjects in their situation regardless of the command. This is dubbed the 'dependence thesis'. It crystallises why the entire theory is a service conception of authority; the authority is only legitimate if it serves the pre-existing reasons of the subject. 10 Further to these three conditions, there are also two key assumptions operating in the background of Raz's conception.
First, the 'reasons' that Raz himself envisions the Service Conception as serving, that is, the pre-existing reasons that we have for action, are all ultimately derived from the good. All practical reasons derive from the respect, promotion and protection of 'value'. Duties, for example, owed to a person in virtue of their personhood arise because they follow from the reason to promote and protect their 9 It also presupposes that one's decisions are informed and effective, not subject to the coercion of another, and made between genuinely diverse and valuable options. See, Raz (1986, pp. 154-155;1995, p. 119, 2006, pp. 1013-1014. See also, Raz (1986, pp. 87, 108). 10 Raz (1986, p. 47): 'The argument for the pre-emption thesis proceeds from another, which I shall call the dependence thesis. It says: all authoritative directives should be based on reasons which already independently apply to the subjects of the directives and are relevant to their action in the circumstances covered by the directive.' valuable 'interests'. 11 We have an obligation to fulfil promises because they promote the value of promising, which derives, in turn, from 'the value of enhancing control over one's life.' 12 We have a reason to be 'autonomous', because it forms 'an aspect of the good life.' 13 Second, Raz is an 'realist' about reasons: that someone has a reason to act is a 'fact', and it may hold true independent of their belief as to whether they have a reason to so act. 14 Beliefs may sometimes be reasons, but in general reasons do not require belief. It is, therefore, (in general) going to be a fact whether or not the normal justification and independence conditions are satisfied regardless of the beliefs of both the authority-agent and the subject.
Finally, whilst the Service Conception is the 'main argument' for the justification of authority, Raz also details three 'subsidiary arguments' that can be used to 'extend' it. 15 The purpose of these three arguments is to recognise the very limited role that consent to and respect for an authority can play as reasons for justifying that authority. In short, in general neither consent nor respect can justify authority that is not otherwise already justified on the grounds of the Service Conception. They are limited to merely an 'expressive' role. There are, however, very 'marginal cases' where the value of the autonomy, loyalty, sense of identification, and/or community achieved through such consent or respect is sufficient to just 'tip the balance' of a regime that would others just fall short of satisfying the Service Conception. 16 In all these cases though, the most important point stands: just as in the Service Conception, authority is only legitimate where one best conforms to reason by following the commands of the authority-agent.

II. THE QUESTION
The Service Conception, therefore, clearly consists in at least three necessary conditions: the normal justification thesis, the indepen-11 Raz (2010, p. 291). 12 Raz (2012, p. 14). 13 Raz (1986, p. 265). 14 It is a 'fact' in the 'extended sense to designate that in virtue of which true of justified statements are true or justified. By ''fact'' is meant simply that which can be designated by the use of the operator ''the fact that…''': Raz (1999a, b, pp. 17-18). 15 Raz (1986, pp. 70, 80). 16 See, Raz (1986Raz ( , pp. 84, 90, 93, 1995. On why such consent can be constitutive of autonomy: Raz (1989Raz ( , p. 1183 dence condition and the dependence condition. Our question is whether Raz presupposes a fourth necessary condition: that the putative subject must believe that the putative authority-agent satisfies the other three conditions.
Let us start with a prima facie interpretation, develop its complications, and then look to an alternative.

III. THE FIRST INTERPRETATION
If we turn to Raz's most recent and supposedly authoritative exposition of the Service Conception -'The Problem of Authority: Revisiting the Service Conception' (2006) -Raz appears to give an explicitly affirmative answer to our question. In what he calls an 'elaboration' (or 'refinement') of the Service Conception he states: The point of being under an authority is that it opens a way of improving one's conformity with reason. One achieves that by conforming to the authority's directives, and (special circumstances apart) one can reliably conform only if one has reliable beliefs regarding who has legitimate authority, and what its directives are. If one cannot have trustworthy beliefs that a certain body meets the conditions for legitimacy, then one's belief in its authority is haphazard, and cannot (again special circumstances apart) be reliable. Therefore, to fulfil its function, the legitimacy of an authority must be knowable to its subjects… [Otherwise] [t]here is no authority over the matter, because to exist authorities must be knowable. 17 This passage appears to state that not merely the subject's belief but 'reliable' belief that the authority agent satisfies the Service Conception's 'conditions of legitimacy' is an additional -fourthnecessary condition of the Service Conception. The idea of a 'reliable belief condition' appears to involve two components: one actual and the other hypothetical. 18 The first, 'actual' component is that a subject must have an actual 'belief' that the authority satisfies the conditions of legitimacy. Presumably, this belief need not be particularly concrete. The subject must simply agree, if asked, that the authority-agent satisfies the 17 Raz (2006Raz ( , pp. 1025Raz ( -1026. 18 Raz nevers explains the precise meaning of 'special circumstances apart'. I shall assume, however, that whatever its intended meaning, its covered in at least one of the interpretations I put forward. normal justification, independence and dependence conditions (and presumably agree that this is sufficient to create a decisive reason for her to act upon the direction of the authority). 19 The second, 'hypothetical' component is that in order for such a belief to be 'reliable' the subject must be able to undertake (but not necessarily have actually undertaken) whatever level of 'inquiry' is reasonable to establish the fact that such conditions of legitimacy are met: In stating this argument I assumed that whenever one can form reliable beliefs that the conditions for legitimacy are met, one can also have knowledge that they are met. … [Something] is knowable if an inquiry of [a kind appropriate to the matter's importance] would yield that knowledge. 20 I presume this means: (a) that the subject would have to conclude their inquiry with, at least, 21 the 'justified belief' that the authority is legitimate; (b) that this belief would have to be sufficiently concrete that they would agree, if asked, that the authority-agent satisfies each of the three other necessary conditions for legitimacy, and (c) that they would have appropriate forms of justification. 22 According to Raz, the rationale for this reliable belief condition is that it is necessary to ensure that authority always remains a way of 'improving one's conformity with reason.' What precisely does this mean?
In another work, Raz distinguishes between 'conformity' and 'compliance' with reason. The former occurs when one acts in a way that is consistent with what one has reason to do. The latter occurs not only when one acts in a way that is consistent with what one has reason to do, but also does so because one understands and is motivated by this reason. 23 Given this definition, however, it is not straightforwardly clear why having a reliable belief that an agent is a legitimate authority is necessary for conformity. For example, we may reliably conform to reason through obeying an authority-agent's command simply because we fear punishment by the authority-agent. In fact, we can be manipulated in all sorts of ways. No belief that the authority-agent is legitimate appears to be necessary.
Does this mean that Raz has, therefore, mistaken conformity for compliance? Is the true legitimate aim of authority, 'improving one's compliance with reason'?
No, this is not correct either. For an agent to comply with reason, in Raz's sense, they must understand and be motivated by the first order reasons at stake. This, however, is clearly not necessary in the case of authority. The entire point of 'exclusionary reasons' is to exclude a subject being motivated by such first order reasons.
Instead, I suggest the following reading: the reliable belief condition is necessary for 'improving one's conformity with reason', if we emphasise the implicit first person act. It is necessary if we are to act with the intention of improving conformity with reason. This appears to be the ultimate aim of the Service Conception and is supported by an earlier passage: In postulating that authorities are legitimate only if their directives enable their subjects to better conform to reason, we see authority for what it is: not a denial of people's capacity for rational action, but simply one device, one method, through the use of which people can achieve the goal (telos) of their capacity for rational action, albeit not through its direct use. This way of understanding matters is reinforced by the fact that in following authority, just as in following advice, or being guided by any of the technical devices, one's ultimate self-reliance is preserved, for it is one's own judgment which directs one to recognize the authority of another, just as it directs one to keep one's promises, follow advice, use technical devices and the like. 24 The (reliable) belief condition, therefore, appears to follow from clarifying what Raz takes to be the purpose of authority under the Service Conception. Authority is not justified because it is a device that improves our conformity with reason, but rather because it is a device that enables us to judge how to act in a manner that improves our conformity with reason. We conform to first order reasons, by complying with the second order reasons provided by authority.
This understanding of the Service Conception, drawn from parts of Raz's most recent work, also coheres with a claim made in one of his earliest account. In his paper 'Legitimate Authority ' (1979), Raz states what he dubs the 'the principle of autonomy': [R]eason never justifies abandoning one's autonomy, that is, one's right and duty to act on one's judgment of what ought to be done, all things considered.
And, in a footnote Raz continues: 'it is clear that this principle of autonomy is not really a moral principle but a principle of rationality.' With respect to this principle, Raz claims that, despite the confusions of others, it does not entail the denial of authority. This is because, correctly understood, authority only entails not acting upon one's judgment of what ought to be done, on the basis of 'first order reasons' (that is, reasons not relating to reasons). However, this is not the totality of possible reasons. There are 'second order reasons' such as exclusionary reasons that give us reasons not to act on certain first order reasons. And, this is precisely what authority offers (that is, exclusionary reasons supporting protected reasons).
Authority, therefore, can -and must -be consistent with the principle of autonomy by reserving for the putative subject the right and duty to act on their judgment of what ought to be done 'all things considered' in the form of valid second order reasons as they see them. Raz does not here expand upon the content of this judgment, but we can now assume that it must, effectively, be the reliable belief condition, that is, one must be satisfied that the putative authority-agent satisfies the three other conditions of legitimacy. 25

IV. THE SECOND INTERPRETATION
Raz's recent claim that authority must be knowable and his early claim that authority must be consistent with principle of autonomy 25 Raz's 'principle of autonomy' is curious. He never justifies it. Nor does he expand upon its implications. Nor does he ever explicitly repeat it again in any further account of the Service Conception as far as I can gather. Further, whilst he dubs it a principle of 'autonomy', he is clearly using the word in a different sense to that he is usually associated with. Usually, as described above, for Raz 'autonomy' denotes a value, that is, the value of (part-) self-creation. It provides a first-order reason for action. It also can be abandoned if outweighed by other reasons. This is precisely the point of the 'independence condition'. The 'autonomy' to which the principle of autonomy refers, however, cannot be abandoned. Further, it is not, strictly speaking, a moral value at all but instead a 'principle of rationality'. both appear to entail that he must assume a (reliable) belief condition as necessary for legitimate authority under the Service Conception. However, other parts of Raz's corpus could not indicate a more different conclusion. I shall focus upon two in particular.
First, in his paper 'Disagreement in Politics', Raz asks what a perfectionist government, established upon true beliefs about the good (and thus reasons for action), owes to citizens who have false beliefs about the good.
Raz accepts that, if the government justifies its authority on the basis of its true beliefs, those citizens who have false beliefs may never be able to accept the government's authority as legitimate. Presumably, this is the case even if they accept Raz's Service Conception as defining what would constitute legitimate authority. The basic problem is that, given deep disagreement about the good, these people cannot agree that the government does indeed 'know best' the reasons that can ground their decisive reasons for action; nor perhaps, appreciate the proper importance of their autonomy. They simply believe, even upon concerted reflection, that the government has got those reasons wrong. Any governmental authority over them, therefore, is inconsistent with such individuals fulfilling both the actual and hypothetical components of the reliable belief condition.
In this situation, in fact a situation Raz argues is inevitable in any Western liberal democracy, he clearly upholds the government's authority. Of course, Raz acknowledges that such a government must cohere with the independence condition and in many circumstances allow those with false beliefs to act upon them, for example, in their daily lives: But it is a far cry from letting people rely on their own views, to denying the legitimacy of the constitution, even though it is based on sound principles, because it is mistakenly disputed by some of its subjects. 26 26 Raz (1998, p. 43). Immediately preceded by: 'Remember that people have a duty to respect their own rationality. That includes the duty not to take the fact that they hold a belief as a reason for anything. There is no reason to think that others must take such beliefs as reasons out of respect for the people who hold them. To be sure, respecting people's rationality is inconsistent with systematically preventing them from relying on their own beliefs.' Footnotes omitted. Elsewhere Raz discusses whether a valid (but not sound) argument that moves from a subject's false beliefs to them concluding that an authority is justified, yields justification: 'But does [this] route yield a justification at all? It relies on the fact that it is possible to derive true conclusions (the principles of legitimate constitution) from false principles. But no such derivation can count as a justification of anything since it relies on falsehood.' 41.
In fact, far from those with false beliefs negating the authority of those running the state: People with principled objection to the government of the day, or some of them, such as Nazis and religious extremists, are an important part of the reason why the rest of us think that governments have authority. They are the people we need a government to control. 27 It follows that, seemingly, that the belief condition must be explicitly rejected: What is normally claimed to be required is that the justification [of the constitution from which a government bases its authority] follows from, or is at least consistent with, views that people already have. If this means that people must be capable of seeing the cogency of the justification without abandoning any of their current views, then this is still too strong and will most likely lead to the conclusion that no Western government is legitimate. 28 This position, seemingly the polar opposite to what we have discussed above, also coheres with that put forward in another paper, 'Government by Consent'. In this paper, Raz reaffirms that, putting aside the subsidiary arguments mentioned above, 29 any legitimate perfectionist government authority must satisfy the three conditions of the Service Conception detailed above. 30 He then asks, what could consent to authority possibly add if these conditions are fulfilled (apart from playing a minor role in the subsidiary arguments for authority)? The answer is 'none': Is not the fact that an authority meets the conditions … sufficient to establish its legitimacy. If so, what room is there for consent? It appears that one can validly consent only to an authority that is legitimate anyway, on independent grounds. … There seems to be nothing that consenting to be governed can do. It imposes no duty and confers no right except those that exist independently anyway. 31 27 Raz (1998, p. 40). 28 Raz (1998, p. 39). 29 The only exception that Raz allows to this position is that consent may play a 'modest role' where it acts an expression of trust in government: such an expression being, in and of itself, valuable. However, trust is merely 'owed to a just and humane government that has legitimate authority in virtue of the conditions of authority in any case.' Only in marginal cases where a government 'nearly' meets these conditions anyway, can it act substantively to 'tip the balance. ' Raz (1995, p. 368). 30 The normal justification thesis, Raz (1995, p. 358); the independence condition, Raz (1995, pp. 366-367); and the dependence condition, putting aside deviant cases: Raz (1995, pp. 358, 359). In this paper, Raz refers to these conditions as the 'conditions of legitimacy'. 31 Raz (1995, 366) [Emphasis added].
Requiring actual consent as a necessary condition for authority would not be the same thing as requiring the fulfilment of the reliable belief condition. The former involves the exercise of a 'normative power over oneself', that is a power to change one's protected reasons. The latter involves merely belief about the whether following another would lead to best conformity with reason. However, if the latter is taken to be a necessary condition for authority, the conceptual gap virtually disappears. This is because the fulfilment of the reliable belief condition entails that the subject must believe that she has best reason to submit to the putative authorityagent. Thus her only rational action, on her own terms, is to decide to obey the putative authority-agent. It may be objected, however, that this still does not constitute consent (or more accurately some form of 'rational consent') because deciding to obey someone -or realising that one should do so -does not necessarily create a duty to do so. This may be true. However, this cannot be Raz's own response, since with the fulfilment of the reliable belief condition, authority exists and thus a duty to obey the authority-agent.
Regardless, even if we can retain some principled conceptual gap between rational consent and the reliable belief condition, there is still obvious 'room' for consent in establishing the legitimacy of a government if the Service Conception includes the reliable belief condition. This obvious role is as evidence that the latter is fulfilled. Consent, assuming it is rational, would be good evidence that the subject actually believes that the normal justification, independence and dependence conditions are fulfilled. However, according to Raz in this discussion, it does not matter what the subject thinks. In fact, 'it is part of the notion of government' that: Governments decide what is best for their subjects and present them with the results as binding conclusions that they are bound to follow. A government … says: ''We are better able to decide how you should act. Our decision is in these laws. You are bound by them and should follow them whether or not you agree with them. '' 32 No matter the precise relationship between consent and the reliable belief condition, Raz's argument that there is no room for the former appears to exclude any room for the latter too.

V. RECONCILIATION?
It is hard to reconcile Raz's remarks within 'Disagreement in Politics' and 'Government by Consent', with his remarks about the need for reliable belief and the paramountcy of the principle of autonomy. At various points, however, he does offer clues as to how this might be achieved. I shall run through various alternatives only to dismiss them and reinforce Raz's choice: the reliable belief condition or perfectionist government, he cannot have both.

A. Dispensing with the 'Actual Belief' Component
At first blush, we might turn to some remarks in 'Disagreement in Politics' that suggests Raz drops the 'actual belief' component of the reliable belief condition, as interpreted above. It is not a necessary condition of authority that a subject actually has a belief that an authority is legitimate. The 'hypothetical' component is, actually, the only relevant condition: Justifications [of the legitimacy of an exercise of authority] are in principle publicly available. There can be contingent reasons why this person or that will find it very difficult to come to realize that the principles are justified. But there is nothing inherently private about justification. Some people are not in a position fully to understand some justification or another, their senses may be impaired, or their powers of understanding limited. For the most part, even they are capable, however, of realizing that the principles or beliefs in question are justified. They would be relying in part on testimony, as we all do in justifying most of our beliefs for which we have any justification. 33 The thought appears to be here, that, although at particular times subjects may not have a belief as to the legitimacy of authority, they could through inquiry (including by using the testimony of others) come to have such a belief, and it would be justified.
The problem with this position is this. The entire force of Raz's comments with respect to politics and disagreement is intended to explain why a true perfectionist government should be able to exercise legitimate authority in the face of deep disagreement, where an political liberal government could not. This is because the Service 33 Raz (1998, pp. 37-38); followed by 'Some people would say that adding 'being capable of being justified to its subjects' to 'being justified' makes a difference, namely, it states that only a justification which can be publicly and comprehensively and explicitly articulated and communicated can lend legitimacy to a regime. But this is a mistake.' Conception is meant to explain why the former government can act upon on the basis of perfectionist reasons in the face of deep disagreement about those reasons.
However, if the Service Conception only applies if the reliable belief condition is satisfied, and the reliable belief condition is only satisfied if, after inquiry, all deep disagreement would disappear, then Raz's entire argument rests on the empirical claim that all deep disagreements are just not that deep after all. However, if this empirical claim were true, then it would just as well provide political liberal governments with the same ability to exercise legitimate authority in the same situations, in the same way, but on the alterative basis that everyone would simply agree, upon inquiry, to the reasons justifying the command. If the empirical claim is true, Raz's theory is redundant. However, if it is false, Raz's theory offers no account of legitimate authority where it is most required.
And the fact of the matter is, clearly, that the empirical claim is false. Indeed, on many issues, such as abortion, or the nature of personal responsibility, or animal rights, or the value of certain cultural traditions, not to speak of the profound general religious and philosophical disagreements, actual, sustained, inquiry on all sides has only entrenched disagreement. This means that any command issued by a perfectionist government, based upon the reasons that it takes to be true, even if they are true, could not possibly satisfy even the amended reliable belief condition suggested. 34 In other words, on the very issues that Raz aims to uniquely empower perfectionist governments with legitimate authority, he fails. We must, therefore, look to a further interpretation of the reliable belief condition to try, once again, to bring some form of internal coherence to Raz's theory.

B. Amending the 'Hypothetical Inquiry' component
A different interpretation of the reliable belief condition is that we not merely abandon the first, 'actual belief' component, but also 34 Sometimes Raz seeks to claim that the extent of actual deep disagreement is much smaller than often claimed. This may be true. However, it is implausible to think that many and very important disagreements shall not remain. One only needs think of those people who think all reasons for action come from the God's command (of whom there are many) and anyone who is an atheist. In this case, they may agree on the rulings that a government may make, but not agree on the reasons for it, even upon extensive inquiry. amend the second 'hypothetical' inquiry component. Under this interpretation, the relevant hypothetical inquiry would not be that which the actual putative subject would undertake if given the chance; but rather that which some form of hypothetical putative subject would undertake.
There is some possible textual support for this interpretation. Raz states: Regarding any justified principle, people of normal capacities are in principle able to understand that it is justified. 35 The claim, here, appears to be, that if people of normal capacities are capable of forming the requisite justified belief after inquiry that an authority is legitimate, then the reliable belief condition is satisfied. It is not necessary, therefore, that people with 'sub-normal' capacities be so capable in order to qualify as subjects. This seems to be supported by this passage: Some people's mental capacities are so limited that it is in principle impossible for them to comprehend the (correct) justification of the principles in question. If that is a problem for the legitimacy of government, it is not solved by taking account of any mistaken ideologies they have, or of their misguided religious beliefs: If their mental capacities are too limited to understand correct justifications, then they are also too limited to understand any plausible incorrect one. The problem they raise, if any, is not one of accommodating disagreement over principles. 36 So, what does Raz mean by 'normal capacities'? Raz appears to equivocate between two very different meanings.
Sometimes, 'normal capacities' appears to track whatever are the capacities of the average person. It is hard to define, precisely what the 'average' person's capacities are, with respect to moral issues. However, loosely we might presume it would correlate with being a member of the majority opinion on the majority of moral issues.
Other times, however, 'normal capacities' appears to track whatever are the capacities of those people who would actually come to realise the true legitimacy of any authority after inquiry. There is no necessary requirement, therefore, that any set of actual subjects would need to actually have such 'normal capacities'. 35 Raz (1998, p. 38). 36 Raz (1998, p. 38). This equivocation gives the appearance of plausibility to this interpretation of the reliable belief condition, but once drawn apart the plausibility disappears.
First, if we take the first possible meaning, then it is not clear why it should be a part of the test for legitimate authority, unless it is driven by an attempt to respect actual, even if unreliable, belief as far as is possible. A majority of people may well be wrong about a moral issue. 37 Why else then should a true perfectionist government, necessarily as a matter of principle, be held hostage to the mistaken view of such a majority, any more than it should be held hostage by the mistaken view of a minority? 38 Such a government may require majority support for instrumental reasons of stability and de facto authority, but that they are merely contingent reasons that speak to whether the normal justification condition holds.
However, if we now turn to the alternative meaning, then the amended reliable belief condition is rendered entirely redundant. As long as the other three necessary conditions of the Service Conception are satisfied, it is hard to see how the reliable belief condition, at least with respect to the authority's moral capacities, would not be trivially fulfilled. 39 No subject need actually be able to satisfy it because, perhaps, they all have sub-normal capacities. 40 What, then, is the point of this amended reliable belief condition? If it does not demand that all subjects, themselves, could know that an authority-agent is legitimate, then what does it add to assure those subjects that at least those who could know the truth, (if they exist) could know? Why not just admit that it is the objective moral fact that the authority-agent satisfies the three other necessary conditions of the Service Conception that actually counts, and stop attempting to insert an additional thinly veiled, trivial 'objectivesubjective' condition? 37 Note, it is irrelevant that, under the right conditions, a majority may get things right, a majority of the time as, for example, argued by Estlund (2008). The reliable belief condition is not there to ensure that the relevant authority actually, in truth, satisfied the other conditions for legitimacy. That is already assumed. Rather, the reliable belief condition is there to ensure that subjects can recognize this fact. 38 Putting aside the contingent factors relating to de facto authority. 39 It may still be the case that no one could possibly understand the truth of some authority agent's legitimacy if, for example, there were certain physical barriers to knowledge. 40 If we think at least some need to have normal capacities so defined, why? Why is the truth itself not enough?

C. A Mysterious Middle Ground
Finally, one may think that Raz is actually seeking to establish some form of middle ground. On this middle ground, all (or virtually all) people are thought to have 'normal' capacities. However, it is also accepted that some of them, if not many of these people with normal capacities, may sometimes be actually incapable of having a justified belief in the true legitimacy of authority-agents. This is because they simply have a mistaken conception of the good. However, we can, so the claim goes, in some sense still attribute such a capability to those mistaken individuals 'in principle'. We might read the following passage as an attempt to do this: Regarding any justified principle, people of normal capacities are in principle able to understand that it is justified. The requirement that the principles on which the constitution is based be justified already includes the requirement that every potential subject of the constitution be in principle capable of understanding that they are justified. 41 The problem is that Raz never makes good on the morally relevant sense in which we can attribute a capability to have a justified belief to someone 'in principle', if they are 'in fact' incapable of so believing after inquiry.
To see the prima facie puzzlement let us take an example. Let us assume that Robert Nozick's political theory was true. Let us also assume, that John Rawls, undertook a significant inquiry, with his own actual beliefs in hand, and that inquiry failed to yield a justified belief that Nozick was right. I would say that Rawls, therefore, was simply incapable of so believing -'in principle' and 'in fact'.
Now we could say that if Rawls had different beliefs when undertaking such an inquiry, then he may have been capable of the relevant justified belief. But that does not seem to me to entail that, although Rawls could not 'in fact' acquire justified belief, he could 'in principle'. Instead, it seems to simply rely upon the morally irrelevant tautology that if Rawls were not Rawls, but someone else, then he might have been able to justifiably believe Nozick. This tautology is morally irrelevant because I could play this trick with any being. If my dog were not a dog, but a libertarian philosopher, then he might 41 Raz (1998, p. 38). be able to justifiably believe Nozick, too. Does my dog, thereby, satisfy the reliable belief condition?
One might reply that Rawls unlike the dog has some 'unrealised capacity' to know (for example, because he is human). But at this point I just no longer have a grip on the morally relevant sense of 'capacity'. I do not understand what it is for me to have a capacity in principle, purely because a lot of beings like me in other ways have that capacity in fact. Perhaps, that lack of capacity is precisely what makes the moral difference between us?
This may all inevitably lead us into the depths of metaphysics of personal identity and radical capacities. Yet, regardless, for instant moral purposes, what is ultimately important is the morally relevant distinction that Raz is seeking to track with the difference between 'in principle capable' and 'not in principle capable.' At this stage, it certainly cannot be Raz's aim to track what a subject would willingly comply with, and that which they would not. Rawls, for example, would not willingly comply with Nozick's theory.
Perhaps one might initially posit that Raz is implicitly aiming to track our (as opposed to the dog's) 'rational capacity' in some sense. Rawls has a rational capacity; a rational capacity is the capacity to follow true reasons for action; therefore, even if Rawls' capacity has here malfunctioned, it could 'in principle' have functioned well. However, this just replaces one mysterious sense of 'capacity' with another. And that sense can only have plausibility because it plays upon another equivocation: between the capacity that all beings, human or animal, have to conform to reason (including by mere manipulation, such as with threats of punishment); and, the capacity that Rawls, by presumption, and other human beings, do not have to comply with reason, that is, to not merely do what reason requires but do so upon because one recognizes its truth. 42 I can only presume that Raz is implicitly aiming to track responsibility in some sense. That is, ceteris paribus we can be held responsible for obeying or not obeying a command that we are 42 Alternatively, one might argue that Raz is seeking to track Reason, that is, that is our 'general capacity reflectively to recognize and respond to reasons ' Raz (2011, p. 86). This capacity may malfunction, for example, when we fail to justifiably believe in the legitimate authority of true perfectionist government. However, in having a capacity to malfunction, the inference is that we have the capacity 'in principle' to function properly. But this just, once again, plays upon a mysterious sense 'capacity' in which only some people are actually capable, after inquiry of 'not malfunctioning', yet all are capable 'in principle' of not malfunctioning. capable of justifiably believing as legitimate 'in principle', but not that which we are 'not in principle' capable of so believing.
This may be true. However, here presented, the metaphysical tail is wagging the dog. If this is the case, we are actually waiting upon a theory of responsibility to clarify Raz's theory of authority, rather than the other way around. We are waiting for a theory of responsibility to clarify, by stipulation, what Raz means by 'the capacity to justifiably believe in principle.' Perhaps, in this case, Raz should replace the 'reliable belief condition' with a more transparently named 'responsibility condition', especially since actual reliable belief would have to be ultimately irrelevant to it? I shall only state at this point, however, that given the prima facie difficulties of attributing responsibility to any agent (dog, human, or famous political philosopher) for something that they are in fact incapable of doing, the onus of proof is well and truly placed upon Raz.

VI. A THIRD INTERPRETATION
Given the inconsistency between our first two interpretations of Raz's position, he appears to be faced with a choice: either the Service Conception includes the reliable belief condition, which ensures respect for the principle of autonomy and excludes the justification of the legitimacy of perfectionist government; or the Service Conception excludes the reliable belief condition, which allows for the justification of the legitimacy of perfectionist government but excludes respect for the principle of autonomy.
There is one further interpretation of Raz's work that would allow him to consistently advocate the legitimacy of perfectionist government, but restricts its authority only to those citizens that fulfil the reliable belief condition. Under this interpretation Raz's theory of 'legitimate government' actually incorporates two different theories of legitimacy.
The first theory is a theory of legitimate authority along the lines described in our first interpretation and which presupposes basic equality as equal fundamental authority. In other words, the government can have no authority over any individuals who do not reliably believe that it satisfies the normal justification, independence and dependence conditions. The second theory is a theory of legitimate coercion that complements the first. This theory justifies coercion, where necessary, with respect to those individuals who are subject to the government's authority in accordance with the first theory. However, it also justifies coercion, where necessary, with respect to those individuals who are not so subject to the government's authority. Most importantly, it justifies coercion against those who do not accept the government's authority because they do not believe that it satisfies the normal justification, dependence and/or independence conditions (although it in fact does).
This interpretation of Raz's position finds support in one passage in particular. 43 Here Raz says: It is important to remember that a government's power [qua right to legitimately coerce] can and normally does quite properly extend to people who do not accept its authority. They are subject to its power partly because those who accept its authority are willing to obey its instructions, even when they affect people who do not accept its authority.
The passage appears to be arguing that the government's right to legitimately coerce extends both over those it has authority over, and those it cannot possibly have authority over because they do not accept its authority. By implication, accepting the authority of a government over oneself is a necessary condition of its having such authority over oneself. This is reinforced by the way Raz then divides the ways a government 'controls and influences people' into three rough but seemingly exhaustive categories. First, some people accept its authority and obey its instructions. Second, the government manipulates the environment in which all its people (those who accept its authority and those who do not) live by constructing roads, digging canals, etc. 'without attempting thereby to exercise authority over them.' And, finally: [A] government controls people by providing remedies for breaches of laws and for the violation of people's rights. People who are not subject to the authority of the law may then obey the rules for prudential reasons, or because even though they have no duty to obey, disobedience will do more harm than good.
So this last category of people appears to be those who cannot be in the first category because they are people that the government does not authority over because they do not accept its authority. However, so the 43 Raz (1986, pp. 102-104). argument goes, the government can have 'power' over them, both to manipulate and coerce, when it satisfies the other three conditions for authority. The fact that 'such action transcends its authority casts no doubt on its justifiability', because in all such matters the government is in 'a position comparable to that of every private individual' who can also manipulate and coerce other to obey their moral duties.
It is difficult to find much textual support for this interpretation beyond this passage. However, it does have the advantage of explaining how Raz can sustain the reliable belief as a necessary condition of authority (coming very close to the political liberal position with respect to government authority) but still advocate the legitimate right of controversial perfectionist government to rule (contrary to the political liberal position with respect to coercive government power).
However, the downside of this interpretation is that Raz must abandon an intuitive relationship between the right of governments to coerce its citizens, and its claim to authority. This intuitive relationship is that the government's right to coerce its citizens primarily exists to back and support its authority. For example, and most obviously, within the criminal system the state punishes citizens precisely because they have to conform to its authoritative directives. It is the criminal's breaking of the law, not their failure to conform to first order reasons independent of the law, that justifies coercion. Now if the third interpretation is correct, then the only citizens to whom this applies will be the procedurally irrational, that is, those who believe that the state is a genuine authority and has given them a decisive reason to not do X, but they do it anyway. Those who reject the authority of the government, by contrast, can act contrary to a law in a procedurally rational way, but cannot be punished for breaking it. The law could never bind them in the first place because they do not fulfil the reliable belief condition. They may be coerced -managed, contained, manipulated and controlled -but not punished. So if a criminal does not have a reliable belief that a law is a law, she cannot break it! VII. CONCLUSION The problem I have in mind is the problem of the possible justification of subjecting one's will to that of another, and of the normative standing of demands to do so. The account of authority that I offered, many years ago, under the title of the service conception of authority, addressed this issue, and assumed that all other problems regarding authority are subsumed under it.-Joseph Raz, 'The Problem of Authority: Revisiting the Service Conception' 44 The problem that Joseph Raz began investigating in his first work on authority in 1979 was about the rationality of subjecting one's will to another. He was asking what type of normative phenomenon legitimate authority had to be, such that one could recognise someone as an authority, and rationally comply with her directives. Framing the 'problem of authority' from this point of view -the point of view of the subject -it is natural that as a necessary condition of it being so rational is that one have a reliable belief that the authority-agent is an authority.
However, whilst Raz began his work by trying to answer this question, he ultimately sought to deliver an answer that was independent of the subject's point of view. He sought to lay out the necessary and sufficient conditions for when X has authority over Y simpliciter. In doing so, in his most definitive texts, he has expressly incorporated the reliable belief condition that was intuitively necessary to answer the original question, into this general theory of authority. Furthermore, when seeking to articulate the most basic rationale of the theory, he has relied upon the condition. To paraphrase the passage cited above, 45 the Service Conception is intended to reveal authority for 'what it is', not a denial of one's rational capacity, but another tool to realise it. And, 'this way of understanding matters' is reinforced by the fact that, with the reliable belief condition, 'one's ultimate self-reliance is preserved, for it is one's own judgment which directs one to recognize the authority of another.' As I have argued, if Raz ultimately wants to maintain this position, then he must give up all pretence to being true defender of perfectionist authority. Depending upon his accompanying theory of legitimate coercion, he will have more in common with political liberals, than the likes of perfectionists such as Hurka,Wall and Enoch. 46 The authority of government will be restricted by the beliefs (false or otherwise) of its citizens, and thus also disagreement between them. 44 Raz (2006Raz ( , p. 1003. 45 See above, Section III. 46 Enoch (2015), Hurka (1993) and Wall (1998). Alternatively, Raz could drop the reliable belief condition. In doing so, the Service Conception would clearly support the extension of a perfectionist government's authority over both those subjects who accept this fact, and those who do not. Further, it could coherently punish both sets of subjects for failing to comply with its authority. In another part of 'The Problem of Authority', Raz reaffirms that the telos of our rational capacity is to conform to reason. But he goes on to point out that we value that capacity primarily, for the end it is meant to achieve, rather than the exercise of the capacity itself: We value the ability to exercise one's judgment and to rely on it in action, but it is a capacity we value because of its purpose, which is, by its very nature, to secure conformity with reason. The point is perfectly general. The value of many of our capacities should not be reduced to the value of their use. But, even where their value also reflects the value of the freedom to use our capacities or not, it depends on the value of their successful use. 47 It would be consistent with this view, that insofar as an authority offers an individual a way to achieve the ultimate purpose of her rational capacity, that is, conformity with reason, then the existence of such authority should not be undermined by the fact that this capacity has misfired in failing to recognize such authority. Following the direction of the authority remains the best way to conform with reason, whether she recognizes it or not. Perhaps, in this way, there lies a principled retreat for Raz to drop the reliable belief condition.
However, it is worth reflecting upon Raz's clear discomfort in ever boldly coming out and holding this position. This is because it is a position that implies that authority can be (perhaps often will be) inconsistent with subjects acting on their own judgement. Raz's old 'principle of autonomy' will have to be abandoned. Instead, subjects will be managed, manipulated, coerced and/or controlled. And this will not necessarily be a small group of the 'irrational', but include anyone disagrees with the putatively true conception of the good upon which the perfectionist government might be based. This may not be a new complaint against perfectionism, but it is one that Raz would need to finally come out and own up to. 47 Raz (2006Raz ( , p. 1017, footnote omitted.

OPEN ACCESS
This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/ by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.