OR TO BE ABLE NOT TO ? THAT IS THE QUESTION . A PROBLEM FOR THE TRANSCENDENTAL ARGUMENT FOR FREE WILL

A type of transcendental argument for libertarian free will maintains that if acting freely requires the availability of alternative possibilities, and determinism holds, then one is not justified in asserting that there is no free will. More precisely: if an agent A is to be justified in asserting a proposition P (e.g. "there is no free will"), then A must also be able to assert not-P. Thus, if A is unable to assert not-P, due to determinism, then A is not justified in asserting P. While such arguments often appeal to principles with wide appeal, such as the principle that ‘ought’ implies ‘can’, they also require a commitment to principles that seem far less compelling, e.g. the principle that ‘ought’ implies ‘able not to’ or the principle that having an obligation entails being responsible. It is argued here that these further principles are dubious, and that it will be difficult to construct a valid transcendental argument without them.


Introduction
Transcendental arguments are typically aimed at refuting sceptical positions. What is distinctive about transcendental arguments is that they do not seek to challenge the sceptic's premises directly. Rather, they might proceed in one of two ways: Firstly, a relatively modest form of transcendental argument may begin with some fact x that is taken to be uncontroversial or obvious (enough so that even the sceptic cannot escape being committed to it) and by arguing that the sceptic's position is inconsistent with x. On this view, the sceptic's argument is not self-refuting, but the sceptic's own commitments cannot be rendered consistent with her conclusion.
Secondly, a more ambitious form of transcendental argument seeks to establish that the sceptic's stance is self-refuting, as opposed to merely being inconsistent with independently inescapable commitments. In this case, the argument will proceed first by identifying some fact x that is argued to be a necessary condition of the very possibility of the sceptic being able to assert her argument, and then by showing that the sceptic's conclusion cannot possibly be true consistent with x. Thus, if the sceptic is able to put forward an argument at all, the argument will be self-refuting. The sceptic essentially proves her own conclusion false the moment she asserts it.
Our aim is to pinpoint and assess some of the key commitments involved in constructing arguments of this sort, with a particular focus on ambitious transcendental arguments in favour of a libertarian stance in the free will debate. We maintain that the success of these arguments depends on whether we can defend not only the compelling principles that typically make these arguments appealing, but also some more dubious principles; those connecting our capacity to make rational choices not only with our ability to do so, but also with our ability to avoid doing so.

Transcendental Argument
Transcendental arguments are traditionally most strongly associated with Kant, who used the method to argue (primarily targeting Hume) that a priori concepts can be legitimately applied to objects of our experience, and to argue (primarily targeting Cartesian scepticism) against idealism (Kant, /1781. Since Kant, the general method has commonly been associated with responses to external world scepticism in epistemology. 1 It's rarer for this argumentative strategy to be invoked in relation to free will, although Kant's own work on free will certainly has echoes of this strategy, and there have been at least a handful of other notable examples. As far back as ancient Greece, Epicurus argues as follows: He who says that all things happen of necessity can hardly find fault with one who denies that all happens by necessity; for on his own theory the argument is voiced by necessity (Epicurus, 1964: fragment XL).
Epicurus does not make it entirely clear why an argument that is voiced by necessity could not be a valid argument for all that. Presumably, the driving assumption is that an argument voiced by necessity is not voiced freely, but he does not clearly spell out why this is taken to undermine the conclusion of the argument. There are, however, a number of ways in which this stance might be motivated.
While not usually regarded as an example of a transcendental argument, Kant's own reasoning in relation to free will in final section of the Groundwork (1997/1785) and in the Critique of Practical Reason (1997Reason ( / 1788 suggests, among other things, that one must presuppose one's own freedom in order to practically act in the pursuit of rational ends. For instance, he argues: Now, one cannot possibly think of a reason that would consciously receive direction from any other quarter with respect to its judgements, since the subject would then attribute the determination of his judgement not to reason but to an impulse. Reason must regard itself as the author of its principles independently of alien influences; consequently, as practical reason or the will of a rational being it must regard itself as free, that is, the will of such being cannot be a will of his own except under the idea of freedom, and such a will must in a practical respect thus be attributed to every rational being (Kant 1997(Kant /1785).
If we must presuppose our own freedom in order to act rationally, then, according to Kant, a commitment to free will is inescapable for any rational being. Moreover, if Kant is right to suppose that we cannot act rationally without presupposing that we have freedom of the sort that would be incompatible with determinism, then it seems to follow that it's also an essential precondition of choosing to argue in favour of a sceptical outlook, at least insofar as one takes oneself to have any practical reason for doing so. 2 While Kant's argument explicitly draws on worries about practical normativity, the Epicurean point could just as easily rest on worries about epistemic normativity. In the latter case, it will be our justification for believing or asserting a conclusion, rather than our justification for acting more broadly, which is taken to commit us to supposing ourselves to be free. Insofar as our status as either practically or theoretically rational entails a certain sort of responsiveness to normative pressures, and insofar as this can be linked with a libertarian understanding of freedom, either might provide a fruitful basis for a suitable transcendental argument for such freedom.
More recently, Lockie (2018) has provided a number of detailed transcendental arguments for libertarianism, which draw on theorising about the relation between freedom, duty, and epistemic normativity, in order to show that any attempt to argue in favour of a deterministic or sceptical position must be self-refuting.
Lockie's argument rests on the idea that freedom is an essential component of epistemic justification. He also draws on the Kantian principle that 'ought' implies 'can' in order to show that determinism poses a serious threat to our capacity to respond intelligibly to epistemic norms. Hence indeterminism is taken to be a necessary prerequisite of anyone being able to justifiably reason to a conclusion -including the conclusion that determinism is true. This requires a broadly internalist and deontological conception of epistemology, according to which the ability to responsibly meet our epistemic duties is a necessary component of epistemic justification (see especially, Lockie 2018, 7-26). If determinism robs us of this ability, then it also robs us of the ability to justify a deterministic conclusion. Hence Lockie's argument forms the basis for an ambitious transcendental argument in favour of libertarian free will.
There is also scope for more modest transcendental arguments, which rest on worries about the practical feasibility of free will scepticism. It has recently been suggested that we ought to interpret Strawson's famous argument in Freedom and Resentment (1962) as a form of transcendental argument for compatibilism (Pereboom 2016;Coates 2017). Essentially, Strawson doubts that we can take free will scepticism seriously, given the commitments that come with the practical perspective forced upon us by our nature as practical agents. It is hardly unintelligible, on this account, to assert that we lack free will, but it may nonetheless be a practical impossibility to wholehearted maintain this view full time.
For the purposes of this discussion, we will put the Strawsonian argument for compatibilism to one side and focus solely on ambitious versions of the transcendental argument for libertarianism; on the question of whether we might have reason to suppose that arguments in favour of determinism are self-refuting in some way. The point is explicit (though underdeveloped) in Epicurus's argument, and is merely hinted at in Kant's reasoning, though it is developed thoroughly and explicitly by Lockie.
Insofar as there is a common theme here, however, the essential claims from which the argument is variously constructed appear to be something like the following: 1. 'Ought' implies 'can' (OIC). 2. Actualism about alternative possibilities: That is, the thesis that determinism rules out the ability to do otherwise; alternative possibilities of the sort required for the ability to do otherwise must be available as things actually are, holding the past and the laws of nature constant (AAP). 3. The ability to do otherwise is a necessary condition of responsibility (PAP).
The Kantian and the Lockiean arguments invoke different further principles pertaining to the sort of normative pressure required for rational action or assertion, while the Epicurean argument leaves this unstated. Though presumably, for Epicurus too, there must be some implicit assumption about the rational requirements for asserting a thesis, where it is supposed that determinism might plausibly preclude us from meeting those requirements. The Kantian principle seems to be something like the following: 4. In order to have any reason to do anything at all, we must have the ability to respond rationally to practical norms (PD).
Let's call this thesis Practical Deontologism. In contrast, the principle that Lockie's argument invokes is explicitly related to epistemic duty: 5. In order to be justified in making any assertion, we must have the ability to respond rationally to epistemic norms (ED).
Lockie calls this thesis Epistemic Deontologism. Either 4 (PD) or 5 (ED) may feasibly be invoked, alongside all or some subset of claims along the lines of 1-3, in an ambitious transcendental argument for libertarian free will. These are all claims that we will be happy to grant, at least for the sake of this discussion. Although they are all controversial, they also each seem to have a fair degree of independent plausibility.
However, we hope to show that in order for any argument of this sort to succeed, there must also be a commitment to one of the following further claims, which we take to be significantly more controversial than the others: 6. 'Ought' implies 'able not to' (OIANT). 7. Duty entails responsibility; no one ought to do something unless they would be responsible for doing it (DER).
Note, that if we take the truth of PAP for granted, these claims essentially become equivalent: The basic idea is that in order to be obligated to do x, we either directly need the ability to refrain from doing x, or we need to be responsible for doing x, where that, in turn, entails (given PAP) an ability to refrain from doing x. Hence what will be needed, in relation to meeting our practical or epistemic obligations, is not merely to be able to, but also to be able not to. That is, for this argumentative strategy to be effective, there are negative and positive preconditions of justifiably acting, asserting, or believing; not only must we be capable of doing what we ought to do, but we must also be capable of not doing what we ought to. It is this aspect of the argument that we take to be problematic.

Determinism and AAP
Following Van Inwagen, we may define determinism as the conjunction of the following two theses: a) For every instant of time, there is a proposition that expresses the state of the world at that instant. b) If p and q are any propositions that express the state of the world at some instants, then the conjunction of p with the laws of nature entails q. 3 If determinism is true, only one future course of events will be possible, consistent with holding fixed the laws of nature and the way that things were in the past. While it might seem intuitive to suppose, at first sight, that the truth of this thesis rules out the ability to do otherwise, there is a great deal of controversy surrounding this point.
According to one reading -we call this the 'actualist' reading 4 -an agent is only able to do otherwise, in the relevant sense, if she is able to do otherwise as things actually stand, holding the past and the laws of nature constant (AAP). On this actualist understanding, determinism rules out alternative possibilities. 5 In contrast, many theorists favour a counterfactual or dispositional reading. On the counterfactual reading, an agent is able to do otherwise if, for instance, she would have done otherwise had she chosen to. 6 On a dispositional reading, an agent could have done otherwise if she would have done otherwise had she been placed in different circumstances. 7 Determinism is consistent with the ability to do otherwise in both of these senses.
While AAP is controversial within the free will debate, it does seem to capture at least one sense of 'able to do otherwise', which goes beyond the conditional and dispositional senses, and which many take to be important for free will. An agent who can act otherwise in the conditional and dispositional senses is one that acts deliberately, acts on the basis of her own choices, and is adequately sensitive to important features of her environment. Many philosophers suppose that this suffices to establish that she acts freely and responsibly. However, while these abilities are almost universally acknowledged to be necessary for moral responsibility, many incompatibilist philosophers have doubts about whether they are sufficient. If the agent is unable to choose otherwise, given the way things actually are, we may worry that she cannot really, in some crucial sense, escape acting the way that she does. E.g. we may worry that she still lacks the ability to act otherwise in a sufficiently robust sense; it may still seem 3 See van Inwagen (1983,65). A similar definition is given in van Inwagen (1975,186). 4 See Elzein and Pernu (2017). 5 Notable defences of the actualist analysis include Campbell (1951), Chisholm (1964), Lehrer (1968), van Inwagen (1983;, and Kane (1999). 6 Notable defences of the counterfactual analysis include Moore (1903), Ayer (1954), Smart (1961), Schlick (1939), Lewis (1981), and Berofsky (2002). 7 Notable defences of the dispositional analysis include Fara (2008), Smith (1997;) Vihvelin (20042014).
unfair to blame her for what she does if she could not actually escape blame, given the way things are. In any case, we will grant AAP for the purposes of this discussion.

Obligation and OIC
The principle that 'ought' implies 'can' (OIC) is popular, 8 but nonetheless remains controversial. 9 There is, however, undoubtedly a great deal of intuitive appeal in the idea that there is something wrong with supposing that demands can be placed on an agent which are impossible for that agent to meet.
In order for this principle to be utilised effectively in any transcendental argument for free will, however, we will need to say something about the sense of 'can' invoked by the principle. Specifically, we will need to suppose that the principle is convincing even granted an actualist reading of 'can'. That is, we must suppose that an agent cannot be obligated to do something unless that agent is able to do it, as things actually stand, holding the past and the laws of nature constant. If we wish to show that determinism undermines our ability to do what we ought to do, in the sense relevant to OIC, then we had better suppose that this pertains to the same sense of 'able to' according to which determinism might plausibly be thought to rob us of the ability to do otherwise.
For the sake of this discussion, we will grant both OIC, and that the sense of 'able to do otherwise' that is relevant to OIC is that invoked by AAP. That is, we will grant that determinism rules out alternative possibilities, and that it does so in a way that entails that we are unable to do otherwise, which, in conjunction with OIC, entails that we cannot be obligated to do otherwise.

Normative Pressures and PAP
There are various ways in which a transcendental argument might run. It may only be necessary to appeal to our capacity to respond to normative pressures, in which case it is not obvious we need to invoke the idea of responsibility at all. But the argument could proceed via a consideration of responsibility if what is taken to be important is not merely the ability to respond to normative pressures, but the ability to be responsible for doing so. In the latter case, the argument may need to make use of PAP: The principle that alternative possibilities are a required for responsibility.
PAP has been under frequent attack at least since Frankfurt's famous attempt to refute the principle (Frankfurt 1969). For present purposes, we will accept PAP, although later we will have reason to consider whether the principle is of central importance to plausible versions of the transcendental argument.
In any case, what any version of the argument will need is some appeal to a normative principle, which bears on when we could have an intelligible basis for making an assertion or for justifying our commitment to a conclusion. Rational justifications for either belief or action must be taken to depend on some sort of ability to respond to normative pressureswhether practical or epistemic. It is this ability that will, if the argument is convincing, be threatened by determinism.

The Basic form of Transcendental Argument
Suppose that we take the principles above to be defensible. This gives us a framework for constructing an ambitious version of the transcendental argument for libertarianism. A simple argument will not rest on PAP, but will instead appeal directly to worries about our ability to respond to normative pressures. This will go as follows: (1) If determinism is true, then nobody is able to do otherwise (from AAP).
(2) If nobody is able to do otherwise, then nobody is able to assert or conclude otherwise (uncontroversial entailment). (3) If nobody is able to assert or conclude otherwise, then nobody ought to assert or conclude otherwise (from OIC). (4) If nobody ought to assert or conclude otherwise, then nobody can have an adequate rational basis to assert or to justifiably conclude otherwise (from either PD or ED).
(5) If determinism is true, then nobody could have an adequate rational basis to assert or justifiably conclude otherwise (from 1-4). (6) If determinism is true, then nobody could have an adequate rational basis for any actual assertion or conclusion (from…?) The problem here is that (6) does not follow from the preceding steps. It certainly doesn't follow from (5) alone. In fact, we only seem entitled to (5). Clearly, however, (5) is a weaker claim than the one needed to establish that any argument for determinism is self-refuting. This would establish that the proponent of determinism cannot have any justification for asserting any alternative conclusion. This looks unproblematic. Insofar as one takes oneself to have a decisive rational basis for asserting a particular conclusion, it follows rather trivially that one cannot be justified in asserting the opposite conclusion instead. To render the determinist's stance problematic, we need a stronger conclusion: That the proponent of the argument for determinism cannot have any rational justification for asserting her actual conclusion.
This could be done either by invoking a principle linking responsibility to duty (DER) alongside PAP, or by invoking the principle that 'ought' implies 'able not to' (OIANT). In either case, we need some means of supposing that the capacity not to fulfil a duty is a necessary condition of being duty-bound, so we end up either directly or indirectly, arriving at something like OIANT.
It's fairly easy to see how the inclusion of OIANT on its own would help to establish a strong enough conclusion: (1) If determinism is true, then nobody is able to do otherwise (from AAP). (2) If nobody is able to do otherwise, then it follows that nobody who makes an assertion or reaches a conclusion could assert or conclude anything other than what they actually do (uncontroversial entailment). (3) If nobody is able to assert or conclude otherwise than they actually do, then nobody ought to assert or conclude as they actually do (from OIANT). (4) If nobody ought to assert or conclude as they actually do, then nobody can have an adequate rational basis to assert or to justifiably conclude as they actually do (from either PD or ED). (5) If determinism is true, then nobody could have an adequate rational basis for any actual assertion or conclusion (from 1-4).
This argument would entail that the determinist would have no rational basis, were determinism true, on which to justify asserting or concluding anything -including the claim that determinism is true.
While the argument could be constructed by appeal OIANT, another route to the same conclusion would arrive at something that entails OIANT, but would commit to it indirectly via PAP and DER, as follows: (1) If determinism is true, then nobody is able to do otherwise (from AAP).
(2) If nobody is able to do otherwise, then nobody can be responsible for anything that they actually do (from PAP). (3) If nobody is responsible for anything they actually do, then nobody can be responsible with respect to the assertions they actually make or the conclusions they actually reach (uncontroversial entailment). (4) If nobody is responsible with respect to the assertions they actually make or the conclusions they actually reach, then nobody ought to make the assertions they make or reach the conclusions that they reach (from DER). (5) If nobody ought to make the assertions that they make or reach the conclusions that they reach, then nobody can have any rational justification for their conclusions or assertions (from PD or ED). (6) If determinism is true, then nobody can have any rational justification for their conclusions or assertions (from 1-5).
The argument may then invoke either OIANT or else PAP alongside DER. The problem, however, is that neither OIANT nor DER are plausible. When there is a compelling practical reason for doing something or a compelling epistemic reason for believing something, we will argue that these pressures are typically independent both of whether we can avoid responding to the pressure and of whether we would be responsible for responding to the pressure. That is, practical and epistemic normative pressures involve the ability to respond to our actual reasons or our actual evidence, and rely neither on our ability to avoid responding to these, nor on whether we would be responsible for responding.
We might suppose that an epistemically rational agent aims to have beliefs that "track" the truth and that a practically rational agent aims to make choices that that "track" their reasons for action. 10 If normative pressures are understood in terms of the obligation to make our assertions and conclusions, as far as possible, track what there is reason to assert or to conclude, it's not at all obvious that either responsibility or the ability to assert or conclude otherwise should be relevant to these pressures at all. While an agent's lack of freedom or responsibility with respect to these pressures may well have an important bearing on whether they can intelligibly be held accountable for their beliefs or assertions, they will not obviously have any parallel bearing on the strength of the agent's reasons for asserting or believing what they do.

The Problem with OIANT
While OIC might seem highly intuitive, OIANT appears to be far less so. While some argue that the two principle are symmetrical in such a way that we ought to accept one so long as we accept the other (e.g. Haji 2002, see especially page 29), it has also been noted that the alleged symmetry is hardly obvious, and unlike OIC, OIANT is rarely seen as similarly axiomatic (Nelkin 2011, 102). Moreover, we might suppose that there is an intuitive rationale for endorsing OIC that simply does not apply to OIANT; we maintain that OIC is plausible because it seems unreasonably demanding to insist that anyone ought to do the impossible. The fact that something is unavoidable, in contrast, certainly does not entail that it would be unreasonably demanding to suppose that someone ought to do it.
Moreover, whether we focus on the epistemic or the practical realm (e.g. on the moral or on the prudential), we will easily find cases in which this principle appears highly counterintuitive. For instance, suppose that you are unable to put your hand into a flame and hold it there for five minutes. Does this really plausibly entail that it's false that you ought to avoid putting your hand in a flame and holding it there for five minutes? Or suppose that you are unable to avoid believing that 1 + 1 = 2. Does this entail that that you lack a strong rational justification for believing that 1 + 1 = 2? Likewise, suppose that you are unable to murder someone in cold blood. Does this plausibly entail that it's false that you ought not to murder here is committed neither to Nozick's externalism about epistemology nor to Fischer and Ravizza's semi-compatibilism about responsibility. In relation to knowledge, the point is that a rational agent aims to have truth-responsive beliefs, where this may be understood as a response to an epistemic duty, consistent with the sort of internalist epistemic deontologism defended by Lockie (2018), say. And while we are suggesting that practical rationality requires the ability to respond to reasons, we are not arguing, as Fischer and Ravizza do, that this suffices for moral responsibility.
anyone cold blood? In all of these cases, it seems plausible to suppose that the answer is no.
The reasons for this cut to what is distinctive about normative pressures. Perhaps you cannot put your hand into a flame and hold it there for five minutes, but this is hardly relevant to the reasons why you ought not to do such a thing. You ought not to do it because you have a very strong prudential interest in avoiding unnecessary pain and injury. This prudential interest will still exist regardless of whether you cannot help but avoid it. Similarly, your reason for believing that 1 + 1 = 2 seems to be just as strong regardless of whether you have the ability to doubt it. The reason is provided by the strength of the mathematical case in favour of concluding that 1 + 1 = 2; that is the strength of the evidence you have on the basis of which to suppose it's true. Likewise, your reasons for not murdering someone in cold blood are based on the fact that it would be morally wrong, not on the fact that you are able to do it.
If we accept OIC, this entails that it would be false that someone who is incapable of avoiding putting their hand in a flame ought not to do so. The obvious rationale is that it cannot be a good idea to do something if that something is literally impossible to do. The practical plausibility of this view appears to be grounded in the fact that it's never practically a good idea to attempt the impossible.
Our point, however, is that it may well be worth attempting the inevitable, especially if there is a causal link between your attempt and your success in that attempt. It may well be inevitable that the moment you realise your hand is in the flame, you retract it fairly quickly. But this doesn't obviously entail that doing so is not also a good idea. You have strong reasons to do it based on the fact that it's in your interests and you are easily capable of doing it. Similarly, if we accept OIC, a person who is incapable of believing that 1 + 1 = 2 is not a person who ought to believe that 1 + 1 = 2. But this does not entail that a person who cannot help but believe it has no reason to believe it.
A plausible form of epistemic deontologism will entail that we have a duty to believe what there is strong evidence for believing, insofar as we are capable of understanding and accurately assessing that evidence. There is no obvious parallel for supposing that we also need the ability to doubt what there is overwhelming evidence to believe. In the case of simple mathematical truths, most of us are likely to find these fairly indubitable.
But it seems odd, to say the least, that we should suppose (in a stark reversal of the Cartesian approach!) that a truth's status as indubitable actually positively undermines our justification for believing it.
One worry may be that we must be committed, in principle, to a strong parallel between 'ought' and 'ought not' requirements. For instance, Lockie argues that because determinism globally denies us the negative, 'irrational', 'unjustified' aspects of any internalist value terms, it removes from us the ability to distinguish and use the positive 'rational', 'justified' aspects of such terms. (If one affects to make no sense of anything being not red, one cannot distinguish and use the predicate red). (Lockie 2018, 182) It is precisely this principle, however, that we take issue with.
Firstly, the parallel between supposing there are no unjustified beliefs or actions, on the one hand, and "affecting to make no sense of redness", on the other, is dubious: The claim is not that we can make no sense of any belief being unjustified, but that if determinism should turn out to be true, then as a matter of fact, nobody is under an obligation not to hold the beliefs they have or under an obligation not to make the assertions that they do. There is an important difference. Consider the idea of non-existence; it is a simple tautology that there exist no things that don't exist. But we can understand the concept of non-existence even if, as a matter of fact, there are no things that don't exist. We are able to make sense of the concept because we are able to think in modal terms; we can contemplate hypothetical scenarios.
There is a great deal of disagreement regarding whether or not determinism is true. Even if we suppose that determinism is true, and we embrace something like OIC, it's not at all obvious that we should be unable to make any sense of the idea that some people ought to believe or assert something different to what they do. This requires that we can imagine a world in which determinism is false, and can think about the obligations we would be under in such a world. This is perfectly consistent with supposing that, as a matter of fact, nobody has such obligations as things actually are. 11 More importantly, the relevant discrimination capacities do not seem to have been located in quite the right place: the normative pressure comes from the strength of the evidence. The relevant ability involves being able to discriminate between strong evidence and weak evidence. A person may well have the ability to discriminate between strong and weak evidence, even if they are not capable of believing anything on the basis of weak evidence, or of doubting something for which there is overwhelmingly strong evidence.

The Problem with DER
We maintain that there is a parallel issue with DER, the principle that in order to have a reason to do or believe something, we would have to be held responsible for doing or believing it.
Again, this appears to misplace the source of the relevant normative pressures. The reasons we have to believe something are dependent on the strength of the evidence in its favour; not on the epistemic agent's blameworthiness or praiseworthiness for so believing. Similarly, the practical reasons we have for acting depend on the moral or prudential case in favour of so acting. Again, where there are strong reasons to do something, these reasons are typically not dependent on whether an agent would be praiseworthy for doing it or blameworthy for failing to do it.
We are not arguing that claims about whether an agent is morally or epistemically blameworthy or praiseworthy are entirely independent of the agent's moral or epistemic reasons: It is clear that if anyone is ever epistemically praiseworthy, a necessary precondition of this is that the agent has good evidence on the basis of which she arrives at her belief. Similarly, if anyone is ever morally or prudentially praiseworthy, a necessary precondition of this is that she had good reasons on the basis of which to act as she did. What we deny, however, is that there is any entailment in the opposite direction: that is, that being praiseworthy is a precondition of having good moral or epistemic reasons. Praiseworthiness, if there is such a thing, depends on there being independent sources of epistemic and practical normativity, not vice-versa.
For one thing, is seems that agents may not be sophisticated enough to be held responsible for their beliefs and actions but may nonetheless have reasons for those beliefs and actions. Consider a five-year-old child who refrains from playing with the loose electrical cables coming out of a live plug socket on the basis that a parent has told her not to. Plausibly, the child is not responsible for her actions since she doesn't really appreciate the reasons why she ought not to play with the electrical cables. But plausibly she ought not to play with them. When her parents tell her that she ought not to touch that live wire, they can hardly be accused of lying to her. She ought not to touch that live wire. The reason why she ought not to touch the wire is certainly not that she will be praiseworthy if she avoids touching it and blameworthy if she touches it (neither of those claims seems plausibly true). In fact, her responsibility doesn't come into it. Rather, she ought not to touch it because she's likely to receive a nasty electric shock if she does.
With respect to epistemic reasons, it seems even more clear that the normative pressures arising from the strength of evidence are not in any way derived from the agent's status as responsible. Suppose the five-yearold works out that 5 x 5 = 25. Perhaps this is quite a difficult calculation for a child or her age and abilities, and it would therefore be unreasonable to suppose that she could be held responsible for successfully working it out. It would certainly be unreasonable to blame her for getting it wrong. None of this seems to have much bearing, however, on why we might suppose that she ought to believe that 5 x 5 = 25. She ought to believe it because it's true and because it's strongly supported by mathematical logic.
Again, the point is that normative pressures arise from facts about what there is evidence to believe and what there is reason to do. These facts do not depend on whether we are responsible. The norms that govern rational belief and behaviour are independent of considerations about whether anyone is responsible for their beliefs and actions.

Conclusion
While there are a number of plausible principles underpinning transcendental arguments for the freedom of the will, they also appear to rest, inevitably, on some principles that we may have good grounds for rejecting. Even if our duties rest on our ability to fulfil them, it is not at all obvious that they similarly rest on any parallel ability not to fulfil them. And while we may have reasons to suppose that our responsibility in relation to our beliefs and actions depends on our reasons, it is far from obvious that there is any dependence in the other direction. It seems, then, that if any transcendental argument in favour of free will is to succeed, it will have to be a significantly more modest form of argument than the sort we have been considering here. It is difficult to see why the determinist could not have good reasons to assert her position without risk of contradiction or self-refutation.