1.1 Inclinations

The standard reading is apparently supported by several of Kant's texts. We can focus on what is perhaps the best-known such text:

This passage suggests the following: (a) the concept of the will of a rational being does not already contain the ‘actual willing’ of an action, or of some kind of action, and (b) CI enriches or adds to this concept of the will in that it ‘connects’ such an actual willing with this concept. If this is right, it is natural to suppose that the will prior to this enriching is an empirically determined capacity governed by norms of instrumental rationality.

The account of empirical willing can be developed along the following lines. The will, quite generally, is the capacity to move oneself to act on some representation in pursuit of the object of that very representation.Footnote ⁶ To find an empirical specification of this capacity in Kant, we need to look at his account of inclinations. For Kant argues that when a suitable encounter with an empirically given object, for example, the eating of an orange, produces in an agent a feeling of pleasure, the experience of this pleasure can – and often does – lead the agent to try to bring about similar encounters with the same kind of object. This process is self-reinforcing, so that the more an agent eats oranges and experiences pleasure, say, the more she seeks out oranges to eat. Thus, pleasure-yielding encounters with given objects can generate in an agent habitual desires for the kinds of objects in question. Kant calls such empirically acquired habitual desires inclinations.Footnote ⁷ And it is an uncontroversial Kantian thesis that we can adopt the object of an inclination as an end, and so act on the basis of the inclination. In such cases, the will is empirically determined in that the motivating factor – the expectation of sensory pleasure – is empirical. And empirical willing consists precisely in that an agent takes the prompting of some inclination as the decisive consideration for adopting an end.

The reading under consideration, however, makes the further assumption that the account of such empirical willing is available independently of any other kind of willing we may be capable of. In particular, it takes there to be a concept of the will in Kant, whereby willing based on inclinations can be made sense of without taking into account necessary prior constraints on such willing. As we will see below (section 2.1), this assumption is untenable.

1.2 Hypothetical Imperatives

In keeping with the Groundwork passage quoted above, the reading construes this will as that of a rational being by introducing rational norms consisting in HIs. We can introduce HIs into the picture as follows. Already implicit in the above account is the conception of the will as a causal power exercised by an agent for the purpose of realizing pleasurable objects. The appeal to the concept of causality here implies nothing more than that an agent moves herself to act in a particular manner and thereby effects a change in the world. The agent is thus the cause of the change in an obvious sense. However, if the agent is to effectively act, she must use appropriate means to bring about what her exercise of causality is in pursuit of. For to pursue an object of choice just is to undertake the means necessary for realizing the object. HIs, according to Kant, are precisely what instruct one to take the necessary means for whatever object one is pursuing.

In a much-discussed passage, Kant suggests that HIs depend on an analytic relation between willing an end and willing the necessary means to that end:

Here, we need not worry about the exact relation between willing an end and willing the corresponding means – whether it is analytic, and if so how. For our purposes, we need simply assume some plausibly close relation between the two. The key point is Kant's suggestion that this relation requires ‘the decisive influence of reason’ in order to be effective in the context of action. In other words, willing an end entails rational requirements instructing the agent to will (or take) the means to her end. And these rational requirements are HIs. Thus, supposing I will to eat an orange, and this requires going to the store to get some oranges, I am bound by an HI instructing me to go to the store (or perhaps to give up the end). If I then fail to go to the store, or to give up my end, I violate an instrumental norm, and thus exhibit irrationality.

Now, given the account so far, it is not immediately clear why the close relation between willing an end and willing the means to this end grounds rational requirements: we are told that reason requires that I will the means to my end, but we have not been told why rational norms enter the picture precisely here.Footnote ⁹ Yet, we might provisionally explain Kant's thought by noting that effectively taking the appropriate means to the ends one wills is commonly taken to be the norm of instrumental rationality. Thus, Kant can be seen as appealing to the common idea that (at least part of) reason's job in the practical sphere is to determine the necessary means to the ends we will, and that insofar as we fail to take these rationally determined means, we exhibit a form of irrationality. Kant's analytic-connection thesis, once spelled out, can then be seen as an account of the norms of instrumental rationality.

The reading under consideration must incorporate some such account of HIs so as to secure the application of reason to empirically determined ends. In keeping with its explanatory strategy, it must further take this account of HIs to be available independently of any other type of rational constraints on choice. That is, on this reading, HIs result from an agent's willing an end whether or not there are other unconditional rational constraints on choice and action. To recap: according to the standard reading, ‘the concept of the will of a rational being’ in the Groundwork passage refers to a capacity that has the following two defining features: (1) it requires an inclination as the determining condition of its exercise; (2) its exercise – the pursuit of pleasurable objects – is subject to HIs, i.e. to norms of instrumental rationality.

Given this concept of the will, CI can be seen as ‘enriching’ it by making possible a new kind of willing. Specifically, CI makes possible acts of choice that are determined not by inclinations but by an agent's cognition of CI itself. Thus, I choose to eat an orange, say, not simply because I will enjoy its taste, but rather because I recognize that eating an orange will help keep me in good health, and that CI requires that I maintain my health. This is a new kind of willing precisely in that it is the cognition of an unconditional rational constraint – rather than the thought of pleasure – that serves as the motivating factor, or the decisive consideration for my choice. Of course for Kant, CI makes possible this new kind of willing because it is a pure rational principle: a necessary constraint on choice that is binding for all rational beings. Hence, CI can also be described as enriching the original concept of the will in that it constitutes a further and stronger rational constraint on the agent's will. Given CI, practical rationality requires not just taking the means to one's ends as prescribed by HIs, but also conforming one's choice of ends to CI. In this manner, the standard reading proposes to incorporate unconditional as well as conditional rational requirements on choice, each applicable independently of the other.

1.3 HIs as Consistency Requirements

Now it may be objected that almost no interpreter of Kant explicitly forwards the sort of reading I have outlined. Yet, this sort of reading is implied by the widely held ‘disjunctive reading’ of HIs.Footnote ¹⁰ As the name suggests, the disjunctive reading takes the requirement embodied in an HI to be disjunctive: it instructs an agent either to will (or take) the necessary means to a given end or to give up the end in question. Thus, the disjunctive reading holds that, given some end, taking the means or giving up the end are two equally valid ways of satisfying an HI. Hence, on this interpretation, instrumental norms are simply consistency requirements. What they demand is that an agent not hold two conflicting or inconsistent attitudes – specifically, that an agent not simultaneously will an end and fail to will the corresponding means.Footnote ¹¹ Any way of avoiding or overcoming this inconsistency then counts as satisfying the instrumental norm.

Yet, the conception of HIs as consistency requirements and their independence of CI imply each other. Consider first that if HIs are independent of CI, they must be consistency requirements. To begin with, it is at best unclear why an inclination as such would give rise to rational requirements, even assuming that the agent pursues the object of the inclination. For there is no intrinsic connection between the promptings of inclination and what an agent is rationally required to do. Yet, presumably, it is the agent's choice – her willing the object of inclination as an end – that gives rise to rational requirements. But if this is the case, the requirements here must be disjunctive. For the mere fact an agent chooses something as her end cannot make it the case that she is rationally required to take the means. After all, the means in question might be morally repugnant.Footnote ¹² Hence, the most that reason can ask here is that the agent be consistent. Thus, so long as HIs are not governed by some other, non-instrumental rational principles – i.e. given that they are independent of CI – they must be disjunctive consistency requirements.

On the other hand – and as has been noted by defenders of the standard reading – if HIs are consistency requirements, they must be independent of CI.Footnote ¹³ For precisely this account of instrumental norms can be endorsed by someone who rejects the very idea of categorical constraints on choice. And many neo-Humean theorists, for example, who would reject CI and indeed any unconditional rational constraints, do explicitly adopt this conception of instrumental norms. This shows that if HIs are disjunctive, instrumental rationality must constitute an autonomous domain, consisting of rational norms that are applicable in just the same way, irrespective of whether CI is a necessary constraint on agents’ choices. It follows that the disjunctive reading of HIs subscribes to exactly the conception of empirical willing and instrumental norms presented above, and this conception is indeed widely held among contemporary readers of Kant.

2.1 Maxims and the Cognition of CI

In the preceding discussion, we started with inclinations as the basis for agents’ choices and actions. To get a better grasp on Kant's account of choice, we must start with the concept of a maxim. A maxim, according to Kant, is a subjective principle of action. ‘Subjective’ here signifies that maxims are principles that an agent in fact acts on. As such, they contrast with practical laws such as CI, which are rational principles that all agents ought to act on. Further, as principles, maxims are the most general rules an agent uses in determining herself to act. Hence, maxims function as the starting points of action: adopting a maxim is the fundamental exercise of agency on the part of an agent. More specific rules of action can then be derived from maxims in the light of relevant circumstances.Footnote ¹⁴

Thus, for example, I might adopt the maxim, ‘I will eat healthy foods’. Given this maxim, and certain further conditions, I may be led to derive the more specific rule of action that I will eat spinach, or even that I will eat spinach tonight. As this example suggests, the point about the generality of maxims is a relative one: maxims can be more or less general. They may also be nested under one another – for example, the maxim that I will regularly eat vegetables might be derived from the maxim that I will eat healthily. The key point, however, is that for Kant, practical reasoning starts with a (relatively) general principle and proceeds by deriving more specific principles and rules from the general principle. And for any such process of reasoning, the relatively general principle that serves as its starting point plays the role of a maxim.Footnote ¹⁵ Further, adopting a maxim and setting oneself an end are necessarily linked: the former entails the latter. Most simply, my maxim may just specify the end I am willing – for example, in the maxim above, eating healthily would be my end. In other cases, for example, when a maxim has the form, ‘I will do A for the sake of E’, maxim-adoption involves more than simply setting oneself an end. Nevertheless, in this as in all other cases, willing an end is necessarily part of adopting a maxim.Footnote ¹⁶

That the structure of practical reasoning proceeds from the general to the particular is clearest and most emphatic in Kant's account of what he calls an agent's disposition (Gesinnung).Footnote ¹⁷ An agent's disposition, for Kant, is her fundamental maxim, one that governs all her other maxims and hence also all her choices and actions. The concept of a single ‘fundamental maxim’ may seem obscure at best. Yet, it shows precisely how Kant's model of practical reasoning and choice is deeply systematic, for an agent's disposition is her most general principle of action, and as such all the other principles and rules she acts on can be seen as deriving from it. What then can this fundamental maxim be? According to Kant, an agent's disposition merely determines the relation between the moral law (or CI) and the agent's happiness. The representation of happiness here consists in an indeterminate conception of the maximal satisfaction of one's sensible desires.Footnote ¹⁸ We can think of it as defined by the maximally compossible set of the agent's inclinations. Given that a disposition determines a relation between this representation of happiness and CI, there are only two possibilities for an agent's disposition. The agent can either adopt a moral disposition, whereby she subordinates her happiness to CI; or, she can adopt an immoral disposition, by subordinating CI to her happiness. Hence, what is in some sense an agent's first or fundamental act of maxim-adoption consists in determining a relation between her sensible desires (generically thought of in the representation of happiness) and the fundamental practical law of reason. This means that this act, and so also every act of maxim-adoption, presupposes and occurs in the face of the agent's cognition of CI.Footnote ¹⁹

Moreover, the basic structure characteristic of an agent's choice of disposition is fully general. That is, we can think of every maxim an agent chooses as her choosing (or reaffirming) her disposition. This is because on every occasion that an agent adopts a maxim, she is in effect faced with the choice of subordinating some sensible desire to a practical law, or vice versa. For instance, suppose that there is a practical law instructing me to preserve my health and so to eat accordingly. Then, on any given occasion where I am choosing what to eat, I can either subordinate considerations of sensible pleasure to the requirement that I eat healthily, or I can choose to satisfy my inclination for unhealthy foods, say, despite the countervailing rational requirement. Hence, each maxim counts as a case of determining a relation between sensible desires and the dictates of reason. In this manner, every time an agent adopts a maxim, she implicitly determines her disposition in that she determines the relation between her sensible desires and practical laws in the context of that maxim. What this shows is that every choice an agent makes – i.e. every act of maxim-adoption – occurs in the face of the agent's cognition of CI (or some other practical law). Hence, an agent's exercise of agency is fully grounded in her cognition of CI as the basic practical law: every choice must presuppose an agent's consciousness of unconditional practical constraints as binding on her will.Footnote ²⁰

This means that there is no room for an independent account of empirical willing of the sort considered above (section 1.1). For according to that account, we can make sense of willing on the basis of inclinations without taking into consideration any categorical requirements – precisely so we can secure instrumental rationality as an autonomous domain. But if the argument here is right, then any willing – any exercise of agency – must take account of CI, and of practical laws more generally. Going back to a point I made above, then, it is indeed possible for an agent to treat an inclination as the decisive consideration for her choice, and so to ‘empirically determine’ her will. What the argument of this section shows, however, is that this is a case of the agent violating CI, or choosing against what she must cognize as a necessarily binding principle for her will. In effect, it counts as a case of the agent choosing an immoral disposition. In other words, what Kant's account of an agent's disposition shows is that there is no possibility of locating a Kantian account of the will of a rational being that does not already presuppose CI as a necessarily binding principle for the will. So also, there is no possibility of starting with the account of empirical willing subject to norms of instrumental rationality and adding on to this categorical norms. Thus, the explanatory strategy the standard reading employs cannot be squared with Kant's account of rational willing.

Now the account developed here seems open to two textual objections, and we need to consider these before developing the account further. First, there is Kant's mention of the ‘most rational finite being’ (1998: 50–1; R 6: 26n).Footnote ²¹ While arguing that CI must be a synthetic a priori proposition, Kant claims that we could conceive of a being that wills maxims, and is perfectly rational in taking the necessary means to its ends, but lacks any cognition of a principle requiring that the form of its maxims be suitable for universal legislation, i.e. of CI. Kant takes this to imply that CI does not apply to this being; i.e. that the being would not be bound by CI. At the same time, this being's characterization as rational suggests that some norms of rationality must apply to it. It is then natural to suppose that these are instrumental norms – this being is ‘most rational’, in other words, precisely because it never violates the HIs it is bound by. The conceptual possibility of the most rational being would then seem to refute the claim that willing a maxim presupposes cognition of CI. In other words, the being in question would seem to lack a disposition in the above sense, since in willing its maxims, it does not determine a relation between CI and its sensible desires.

Now what Kant claims is in fact a bit weaker – his claim is strictly that we cannot conceptually rule out the possibility of such a being, of a composition of capacities whereby the possibility of maxim-formation does not itself entail cognition of the fundamental constraint on this activity. But this is simply another way of saying that CI is a synthetic a priori proposition – that analysis of the concept of reason, or even of a rational representation like a maxim cannot by itself yield cognition of CI. As Kant elsewhere puts the point, the cognition of CI is an irreducible ‘fact of reason’.Footnote ²² This by itself need have no implications for the universality and necessity of the latter cognition. Indeed, the designation of CI as synthetic a priori implies just the contrary, namely, that it is universally and necessarily cognized by agents insofar as they will maxims. Insofar as Kant's discussion of the most rational being suggests something stronger, namely, that the synthetic a priority of CI does open up space for the possibility that its cognition is not universal, it should be resisted. The most plausible way to make sense of Kant's position as a whole – in particular the designation of CI as synthetic a priori – is to conceive of the cognition of CI, and hence the applicability of the notion of a disposition, as being explanatorily irreducible and yet strictly universal.Footnote ²³

Secondly, there is the question of how to read the Groundwork passage we started with, and which seemed to strongly suggest the standard reading. Let us consider the passage closely. As in the case of the most rational being, Kant's point here is to argue that CI is a synthetic a priori proposition. Hence, Kant first claims that CI ‘does not analytically derive the willing of an action from some other willing already presupposed (for we possess no such perfect will)’. The ‘willing of an action’ is plausibly construed to be some action required by CI – the repayment of a loan, say. Such a willing, Kant claims, cannot in our case be derived from some prior willing which we can presuppose – though it could be so derived if we possessed perfect wills. Why this parenthetical reference? Well, a perfect will would be one that never violated CI: it would always – perhaps necessarily – will for itself a moral disposition. Kant's point then is this: if we could presuppose the willing of a moral disposition, we could analytically derive the willing of an action such as repaying a loan. In our case, on the other hand, the choice of an immoral disposition is a standing possibility, making the above presupposition impossible. Hence, CI must ‘connect the willing of an action immediately with the concept of the will of a rational being’. That is, cognition of CI must directly govern one's choice to repay the loan, and this choice will (re)affirm one's moral disposition (rather than presupposing it). Just as above, there is no suggestion here of a possible type of choice or willing that is not governed by CI. Rather, Kant's claim, again as above, is that cognition of CI as a necessary, synthetic a priori constraint is an irreducible ‘fact’ which is to directly govern our choices.Footnote ²⁴

2.2 Consistency vs. Strict Dependence

What the last subsection shows is that in Kant's system there is no possibility of a choice or a willing not subject to CI as a necessary constraint. Hence, there is no possibility – as in the standard view – of starting with a conception of empirical willing governed by HIs and then adding on CI as a stronger constraint. Nevertheless, this does not yet settle the question of whether particular HIs are strictly dependent on CI: whether they follow from all ends, whether licensed by CI or not, or whether they are applicable when and only when an agent wills a permissible end. For the following sort of hybrid view still seems possible. Suppose we accept that instrumental rationality is not a fully autonomous domain – that the applicability of HIs in general requires cognition of CI on the part of agents willing maxims. Yet, we continue to think of HIs as disjunctive consistency requirements. That is, we take CI as a necessary background constraint on choice, and posit a further set of requirements generated simply by agents making choices, these being HIs construed as disjunctive consistency requirements. Hence, although instrumental rationality is not a fully autonomous domain, HIs apply independently of CI in the sense that they apply in the same way irrespective of whether an agent's choice conforms to CI.Footnote ²⁵

It is instructive here to consider Christine Korsgaard's influential argument in ‘The Normativity of Instrumental Reason’. Korsgaard argues for a conclusion similar to that of section 2.1, namely that norms of instrumental rationality in general must presuppose CI.Footnote ²⁶ Yet, on the question of whether HIs are consistency requirements or whether they are strictly dependent on CI, Korsgaard's position is at best unclear. Generally, Korsgaard is read as holding something like the hybrid view.Footnote ²⁷ And several passages in Korsgaard do strongly suggest this reading. For instance, Korsgaard suggests in a footnote that she largely agrees with Hill's account of instrumental norms, which squarely defends treating HIs as disjunctive consistency requirements. Further, towards the end of the paper, Korsgaard acknowledges the possibility of a ‘heroic existentialist’, who commits herself to some end, regardless of its content, and thus generates HIs she is bound by. Yet, if HIs can follow from this sort of existentialist choice, they cannot be strictly dependent on CI.Footnote ²⁸

These passages, while not conclusive, strongly suggest that Korsgaard does not forward a settled view on whether HIs are disjunctive consistency requirements or whether they are strictly dependent on CI. Of course, the question of the form of specific HIs and the conditions of their applicability is not an explicit concern of Korsgaard's, so it is unsurprising that it is hard to find a clear answer to this question in her paper.

Yet, given the recent interest in questions of scope and in whether instrumental norms should be understood as consistency requirements, it is useful to explicitly consider what the Kantian account implies for this question. More importantly, to properly grasp the conditionality of HIs in Kant's picture, it is essential to see that HIs are indeed strictly dependent on CI. This is what I propose to do now – showing first that Kant could not take HIs to be consistency requirements, and then, in section 2.3, outlining the conception of HIs as narrow-scope requirements that follow from, and only from, the willing of permissible ends.

Support for the view that HIs must be consistency requirements can be drawn from certain passages in Kant and from certain intuitions about the ascription of instrumental irrationality. Considering the textual support first, the reading of HIs as consistency requirements is most strongly suggested by Kant's discussion of ‘imperatives of skill’ in the Groundwork. Kant writes of such ‘imperatives’: ‘Whether the end is rational and good is not at all the question here, but only what one must do in order to attain it’ (1997b: 26; G 4: 415). And he goes on to suggest that such ‘precepts’ are ‘of equal worth’ to a physician trying to cure as to a poisoner trying to kill. Now Kant does not explicitly say that the poisoner is bound by an HI, but it might be thought that insofar as imperatives of skill apply to him, as the passage suggests, he must after all be bound by relevant HIs. Since, however, the poisoner would presumably not be rationally required to take the necessary means to his end – to inject the poison, say – the HIs he is bound by must be disjunctive ones. They must, in other words, be consistency requirements.

Yet, despite the mention of ‘imperatives’, it is far from clear that Kant's claim here is that the doctor and the poisoner are bound by structurally similar HIs, i.e. requirements they must fulfil on pain of irrationality. In this passage, Kant is discussing ‘problematic practical principles’ – ‘oughts’ that may apply were an agent to adopt a given end. Hence, the topic of discussion is not the actual HIs that a given agent is bound by. Kant's point is rather that it is a sound policy on an agent's part to develop the skills that can help her accomplish the ends she may in the future adopt. So, for example, it is ceteris paribus good for agents to know how the human body reacts to various substances – knowledge they can then use to bring about their ends. This, however, does not imply that an agent who wills to poison someone is indeed bound by an HI, whether disjunctive or otherwise.

More importantly, when Kant revisits this topic in the second Critique, he designates such precepts ‘rules of skill’ and denies that they are practical principles at all:

Yet, if such rules of skill are theoretical principles – instances of theoretical knowledge – that in general terms tell agents what they must do if they are to bring about their possible ends, they cannot be constitutive of requirements of practical reason. They cannot, that is, be instrumental norms instructing agents what they ought to do on pain of irrationality.Footnote ³⁰ Hence, on the most plausible reading of the Groundwork passage, the ‘imperatives’ in question should not be construed as disjunctive HIs.

Further, there is good reason to think that Kant would reject the thought that practical reason in general requires consistency. To see this, consider a more systematic argument for treating HIs as consistency requirements, which we can develop by reflection on our intuitions about the ascription of instrumental irrationality. Suppose an agent sets herself an end that violates CI – say, she wills to rob her neighbour while the neighbour is away on vacation. Suppose further that a necessary means to this end is that the agent disable her neighbour's alarm, and that while continuing to hold on to the end, the agent fails to will or take these necessary means. The agent is here being inconsistent, and there is a strong inclination to say that she is instrumentally irrational just in virtue of being inconsistent. If this is right – if the would-be burglar is indeed exhibiting instrumental irrationality – it must be because she is violating an HI. And again, the HI in question must then be the disjunctive requirement that she either take (or will) the means of disabling the alarm or give up the end of burglarizing. Thus, accommodating intuitions about instrumental irrationality may incline one to construe HIs as disjunctive consistency requirements.

Yet, this is not enough to justify treating HIs as disjunctive. Per hypothesis, what the would-be burglar gains by disabling the alarm is consistency – this is precisely how she can avoid instrumental irrationality. Yet, there is no reason to think that practical reason would require mere consistency. To see this, consider Kant's famous insistence that a good will is the only thing good without qualification – every other aspect of an agent's make up can be good in some circumstances, but is not necessarily so.Footnote ³¹ This plausibly applies to consistency as much as to anything else, so that consistency is not a general requirement of practical reason. Yet, if consistency is not a general demand of reason, then reflection on cases like the would-be burglar's does not give us any reason to construe HIs as disjunctive requirements that apply irrespective of the character of one's ends.

Further reflection on the case of the would-be burglar bears this out. For the necessary rational demand in this case would be that the agent give up the end in question, and so also that she not bring about her end. Consistency would imply that ceteris paribus the agent does bring about her end. But this is hardly a rationally satisfactory outcome. In terms of what reason requires, then, the would-be burglar is no better off if she is instrumentally consistent (and given that she brings about her end, she may well be worse off). Hence, instrumental consistency is no virtue on the would-be burglar's part, and it is at best difficult to see why reason would require that the agent be so consistent. Yet, if there is no rational requirement prescribing consistency, the would-be burglar is indeed not guilty of instrumental irrationality. In general, then, agents who will ends contrary to reason are not instrumentally irrational if they fail to take the means to their ends.Footnote ³² Hence, given that CI is a necessary and unconditional constraint on the will, HIs cannot be consistency requirements that bind agents irrespective of their ends.

2.3 Hypothetical Imperatives (again)

The last two subsections show that on Kant's view (1) instrumental rationality cannot be an autonomous domain, and (2) HIs cannot be consistency requirements. Given the argument of section 1.3, this means that HIs must be narrow-scope requirements that are strictly dependent on CI, i.e. that they apply when and only when an agent conforms her choices to CI. Here, I will delineate this account of HIs, focusing on their status as conditional requirements generated by the willing of permissible ends.

Consider first that HIs, as requirements pertaining to the willing of means, come into the picture after the adoption of maxims. Given that maxims are general principles of action, HIs can be seen as pertaining to the more specific rules of action that can be derived from a maxim. For instance, suppose that my eating spinach for dinner tonight is necessary (or effective) given my maxim of eating healthily. I would then be bound by the HI, ‘I ought to eat spinach for dinner’. Now the force of this imperative derives from my prior commitment to eating healthily: there is no independent force to the imperative requiring that I eat spinach tonight. Similarly, suppose one of my maxims is that I will regularly watch independent and art-house films. Quite plausibly, such a maxim would be perfectly permissible but not in any way required by a practical law. Suppose further that, given my maxim (and other relevant conditions), I have compelling reason to attend the single screening of a given film. Then, I am plausibly bound by the HI, ‘I ought to attend the screening of X’. It is evident that the binding force of this instrumental norm is completely derivative on my prior commitment to watching certain sorts of films. Generally, then, the bindingness of HIs follows from the prior bindingness of the maxims from which the HIs are respectively derived. Minus the bindingness of the maxims, there is no way to make sense of HIs as rational requirements.

The question, then, is what exactly makes maxims binding, and whether there could be maxims that are not binding in the requisite manner? And the answer must be that the binding force of maxims depends on the practical laws that are to function as necessary constraints on choice. That is, it is precisely because my film-watching maxim is rationally licensed that it is binding and can in turn generate HIs. Now it might seem odd that willing a merely permissible end can give rise to rational requirements that I undertake the necessary means to this end. Yet, recall that every choice reaffirms an agent's disposition: since every choice must be made in the face of the agent's cognition of CI, it must reaffirm the relation between CI and happiness for the agent. Hence, the choice of permissible ends must be made in consciousness of the fact that the end in question does not violate CI. Indeed, for an agent's end to be genuinely permissible, it must be chosen by the agent as permissible, or on the ground of the recognition of its rational licence. Absent the consciousness of the rational licence, it is not evident that we should even call the choice permissible – at best, the permissibility of such a choice would be purely accidental.

Hence, the willing of permissible maxims and ends reflects and reaffirms an agent's moral disposition. And this means that ‘merely permissible’ maxims and ends are indeed rationally determined: the decisive consideration for the choice is conformity of the maxim with CI. That said, there is still a perfectly intelligible distinction between permissible and obligatory maxims – an agent is rationally licensed to revoke her willing of one of the former, though not of the latter. Nevertheless, insofar as an agent wills – and continues to be committed to – a permissible maxim, she is required to take the necessary means to bring about the end specified in the maxim. And the basis of this requirement is precisely the role of the agent's cognition of CI in her choice of maxim. That is, precisely because the agent cognizes CI as a binding constraint on her choice, and chooses accordingly, her choice generates conditional requirements to take the means to her end. On the other hand, if an agent chooses a maxim that violates CI, or is impermissible, the maxim itself lacks binding force. Instead, the agent, necessarily bound by CI, is required to give up the maxim in question. Given this, the choice of such a maxim also cannot generate HIs. Indeed, it is because impermissible maxims lack binding force that no HI follows from them.

Hence, the bindingness of HIs – that they are rational requirements – derives from the bindingness of CI. It is because an agent's permissible ends reflect a moral disposition that she is required to take the means to these ends. And this is how HIs can function as conditional requirements of practical reason. They are, like CI and other practical laws, genuine requirements of practical reason in that they prescribe the exercise of agency in determinate ways (rather than requiring mere consistency). For instance, in the example above, I am rationally bound to exercise my agency so as to attend the film screening. Hence, HIs require of an agent that she will and act in the determinate ways specified by the imperatives themselves. Insofar as an agent fails to exercise her agency in the manner specified by the HI – i.e. insofar as she fails to take some necessary means to her end – she exhibits instrumental irrationality.Footnote ³³ At the same time, HIs, unlike practical laws, are conditional requirements: they bind agents only on the assumption that certain prior conditions have been satisfied. Fundamentally, the bindingness of instrumental norms is conditional on an agent's conforming her choice of maxims to CI (and other practical laws). Yet, given that the choice of permissible ends reflects a moral disposition, we can equally say that HIs are conditional on an agent's choosing permissible ends. Hence, an agent's exercise of agency in willing a permissible end binds the agent to exercise her agency so as to will and take the means to this end. And it is just this rational demand – that the agent bring about the end she has willed by taking the necessary means – that HIs express.