Free Will and Epistemology

10.1 Compatibilism and the transcendental argument

Given that this ‘third way’ – of self-determinism – establishes a coherent alternative to the determinism/indeterminism dilemma, the question remains as to why it should be preferred to soft determinism. One venerable way of answering this question is to turn the tables on the conceptual objections of the classic compatibilists: on their picture of things, the notion of a power to determine not just one’s acts, but one’s will, is incoherent. On a self-determinist picture of things, it is coherent. If, then, any example of this power can be shown to be actual, this refutes the conception that entails it is incoherent.

Manifestly, it does make sense to talk of ‘choosing to choose’ and ‘desiring to desire’. An obese person wants to have a taste for healthier foods: he doesn’t desire these foods now, but he wants to desire them. A timid person wants to be more outgoing – she wishes she were more extravert – where an extravert is just by definition someone with different social appetites to the introvert. An obsessive–compulsive has an overwhelming desire, will, compulsion to wash her hands until these be red-raw. She has the great desire to be free of this desire, however. It is on the basis of this that sense can be made of her going to a behaviour therapist. She chooses to will differently. Since these attempts to change our will are sometimes successful, the claim that we cannot ever choose our will is false, and, a fortiori, the framework of assumption which entails that it is senseless must be in error. Conative ‘voluntarism’ (if you must) is actual – we see it in our own and others’ lives.

As an argumentative response to compatibilism, this would have been highly effective until about four and a half decades ago. There has, though, of course been a revolution in compatibilist thinking from Frankfurt (1971) onwards, and few philosophers, whether compatibilist or incompatibilist, now doubt that talk of higher-order cognition and conation – of the kind implacably prohibited by the classic Hobbes–Locke tradition – is in fact entirely legitimate. The challenge presented by such higher-order cognition and conation led to two main responses from an evolving compatibilist tradition: the older, ‘strongest motive’ opposition, which maintained that such higher-order conation was in fact a logical or psychological construction built out of an associative sum of first-order conation; and the newer, ‘quasi-realist’ (mesh account) response of the last forty-five years. The former has been dealt with in the previous chapter. The latter will be considered now.

10.2 The quasi-realist response

From the early 1970s, over a period of just a few years, compatibilists en masse abandoned as indefensible the line that they, after Hobbes, had resolutely tried to hold for more than 300 years, finally surrendering what those unencumbered by their theoretical commitments had seen as obvious throughout: that higher-order choosing – Reid’s determination of one’s will – was not only conceivable, but actual. The new line that most, though not all, retrenched to, and have in their different ways defended since, is that this is not sufficient of itself to establish a freedom of self-determination: a determination of the self by the self. For to suppose that these higher-order choices were determined not by the self as such, but by another Hobbesian conative item, itself determined by natural law, would be no more than an iterative addition to the soft determinist’s account. This, of course, is the position of Frankfurt, and, with significant differences of detail, of those who succeeded him with other mesh accounts: the first major development of the compatibilist position since Hobbes. Moral responsibility requires personhood, where a person just is a creature with second-order desires (Frankfurt) – or, mutatis mutandis, such mental contents as satisfy the details of these other accounts. These desires to desire are, however, held by this tradition to be themselves just determined mental items in turn. Why, in the face of this compatibilist retrenchment, do we have to suppose that they must be self-determined instead?

As matters were left in Chapter 8, determinism as such was presented with transcendental arguments in three different forms. The obvious response to such arguments was, as indicated, to soften determinism such that this thesis no longer possessed the deflationary characteristics capable of reflexively undermining it. Classic compatibilism, though more or less obsolescent as a position, offers us a paradigm of what such a softening might look like; and how, structurally, a response to these arguments might go. Successor soft determinist theories will attempt to pick up where it left off. It cannot be expected that we here consider the prospects for each such candidate theory in turn (to do this, even a book-length treatment would probably not suffice); but we may see how, schematically, one such category of theory has poor prospects of meeting the reflexive, justificatory demands of these kinds of arguments, then to develop some general morals therefrom.

Hard determinism is an error theory of human freedom. The hard determinist agrees with the libertarian that the meaning of a claim to freedom is that the person exercises a power of determination over her actions and ratiocination, and that the truth of this claim is incompatible with natural law also having this power. Where the hard determinist disagrees with the libertarian is over the question as to whether we are ever able to determine an action, ratiocination φ. That is, hard determinism is committed to 1, below, being always false:

1. Person P has a power over φ (for some action/ratiocination/argument φ).

This invites us to make the claim, ratiocination, argument, for hard determinism or its consequences, a substitution instance of φ in the putatively erroneous 1. And that, on a deontic leeway assumption of the kind defended in this work, leads to the conclusion that hard determinism is without epistemic justification. It leads also to the Epicurean frontispiece point from Chapter 8: that the determinist is powerless to take his libertarian opponents as being unjustified, since they too are powerless to do or be convinced otherwise. The libertarian is not similarly hamstrung.

So classical soft determinism provides a revisionary account of what the power to determine an action consists in. On this view, 1 above is not false. 1 just means whatever the details of the species of compatibilism in question require it to mean – here, for a neo-Hobbesian dummy account:

2. Person P has not been thwarted in carrying out her action, ratiocination, argument to φ, in accordance with her decision, intention or desire to φ, where, however, this decision, intention or desire was itself wholly determined by natural law.

This, the neo-Hobbesian will maintain, is what saying that the person has the power to act, ratiocinate, argue means.

Notice that this soft determinism opposes hard determinism no less than it opposes libertarianism, and has to do so, on pain of being unable to resist the transcendental argument against the former position. It does not, however, collapse into any kind of ‘hard libertarianism’, since 2 clearly holds that the power to choose is restricted to first-order choice – a freedom from constraint in acting on our determined mental items. There is no freedom to choose these items: the person has no power to determine them – they are determined by natural law, externally conceived.

That is, soft determinism cannot accommodate a statement such as 3:

3. Person P really has a power over φ (for some action/ratiocination/argument φ).

Where ‘really’, to avoid question-begging, is (for this dummy account) cashed out as:

3.1. Person P has a power over her decision, intention or desire φ′ to φ.

Ordinary soft determinism must hold this to be false.

However, as noted, it is easily established that such talk as this makes sense – that, pace Hobbes, Reid’s claim to a ‘power over the determination of his own will’ is not some kind of conceptual error. Further, it is easily established that, as a matter of fact, instances of this power over our will are actual – so soft determinism stands refuted. Here, however, soft determinism is already established as being reflexively untenable; for, if all power to determine our mental items is outside of our control, a requirement of internalist justification is denied us. Ratiocination, argument, deliberative judgement – whether in private or public – requires, contra Hobbes, that the person be responsible for what he called their ‘trayne of thought’. Let us suppose these trains of thought are (as on the philosophy of mind that has historically accompanied the soft determinist view) contiguous successions of mental items connected by frequency of co-occurrence – though some more substantial principle of connection could be substituted here. The mental items themselves, those items that are so connected, are just those things over whose occurrence we supposedly have no power. But if the person has thus no power to determine these, their individual mental items, a fortiori their order of succession and development, then the person has no responsibility for their train of thought.1 The freedom required for rational justification to be present cannot be merely the ‘freedom’ to have an externally determined train of private thought or public argument come forth or be emitted.

That is, the claim, ratiocination, argument for soft determinism or its consequences may be made a substitution instance of φ (φ′) in the putatively erroneous 3 (3.1). Soft determinists must maintain that neither they nor their opponents (incompatibilist determinists and libertarians alike) really have the power to determine their arguments, ratiocinations and so on. This reprises the defensive motivation behind the soft determinists’ position – their need to avoid the re-emergence of the transcendental argument that undermines hard determinism. The quasi-realism we are envisaging will hold that 3/3.1 can be accommodated by an extension of the determinist’s position – that the determinist need not maintain 3/3.1 is in error. That is:

3. Person P really has a power over φ (for some action/ratiocination/argument φ).

is not false. It just means:

4. Person P has not been thwarted in carrying out her decision, intention or desire φ′, to form a decision, intention or desire to φ, where, however, that decision to decide, desire to desire, intention to intend, is itself wholly determined by natural law.

To describe this developing series, we may borrow some terminology from modern neo-Humean metaethics. Hard determinism, as noted, may be seen as a direct error theory of human freedom. Soft determinism (classic compatibilism) is the simplest form of non-literal semantics for ‘freedom talk’. That is, within a world wholly bereft of real freedom, an ersatz simulacrum of such is constructed: to be free is to have one’s (determined) desires be unfrustrated by external constraint, to be conatively potent is to have one’s (determined) efforts unobstructed, and so on for all the other cases. Finally, a kind of ‘quasi-realism’ may be developed – the open-ended iteration of this approach to accommodate cases that cannot be handled by the simpler approach. Notice that this quasi-realist compatibilism opposes the earlier, more familiar soft-determinism out of which it grew, no less than this latter position opposed the hard determinism out of which it grew, and in each case for the same reason. Soft determinism had to oppose hard determinism to resist the transcendental argument that could be developed out of the deflationary commitments that followed from this position. The (softer still) quasi-realist position of Frankfurt and those other accounts that followed his must then oppose ordinary soft determinism in turn – to avoid the re-formed transcendental argument that can develop out of this latter position. That is, any softening of determinism is, again, no concession on the part of the determinist.

As was the case for ordinary soft determinism, this envisaged quasi-realism is not supposed to collapse into any kind of hard libertarianism, since it holds that the power to choose is restricted to (now) only first- and second-order choice: We have a freedom from constraint in acting on our determined mental items, where some of these determined mental items are now items that determine our other mental items, and thus ultimately our train of thought, our arguments and deliberative judgement. There is, though, no freedom to choose the items with this executive power (these second-order cognitions and conations) – they are determined by natural law, externally conceived. They in turn determine all else – all argument, ratiocination, action and judgement. To appropriate some terminology from Kane (1996), we have the appearance of ‘will-setting’ on this view, but no ‘ultimate responsibility’2 for that will-setting – as the will is not set, is not formed, by the self. Our will-setting takes place by determined prior mental events – or indeed, not even mental events.

It is clear enough how the counterarguments to this position will in turn proceed, and it is obvious how the quasi-realist position will respond. A re-formed transcendental argument can be levelled against this position, and one must expect its defenders to respond to any such argument by a softer-still determinism, a yet more sophisticated, third-order quasi-realism.

Throughout this extended discussion, we have maintained, with the soft determinist tradition, the fiction that a conation will consist in an actual, separable item, à la Hobbes, Sechenov, etc., in the mind/brain – not an assumption the self-determinist will typically accept. Still, operating on this assumption, it is clear that any such regress of higher-order conation could not be actually infinite. Often, it will not actually be even begun – mostly we just choose; we don’t make a discrete, actual choice to choose (cf. codicil to Chapter 9: Section 9.8.4). Given this fact, isn’t it unfair to press the compatibilist as above – can’t their response be that at one level or other the chain of conative determinants comes to an end?

Suppose it does – the question is, with what? Does the determination of action, ratiocination and the like end with a determined component of the mind – an unchosen, ontically real item in the mind/brain: a desire, ‘volition’, act of will, etc., that natural law and the Big Bang necessitated (or, for that matter, that blind chance brought to pass)? Or does it end with the self, the agent, the person, their mind as a whole (in the ethics case, their conscience)? What determines the last conation? This ‘last’ conation may be the first – as for the case of ordinary first-order conation – or it may be the nth. But the question cannot be indefinitely deferred: whichever it is, if it is determined by natural law, externally conceived, then the agent no more satisfies the deontic requirements for rational justification than they would on the picture sanctioned by the simplest form of hard determinism.

The first and most plain argument given was that if we did not have a power to determine our actions and ratiocinations, we could not satisfy internalism’s requirements for justification. That is, if we, ourselves, did not determine them, as source, but a finite set of externally conceived natural laws did, then we would not be responsible.

For to say that what depends upon the will is in a man’s power, but the will is not in his power, is to say that the end is in his power, but the means necessary to that end are not in his power, which is a contradiction. (Reid 1969: 266, Essay IV Bk 1)

So long as ‘nothing taketh beginning from itself but from the action of some other immediate agent without itself’, a soft determinism, however sophisticated, will no more be able to satisfy the requirements of reflexive epistemic justification than will the plainest hard determinism. Iterating the causal chain that gives rise to the action or ratiocination, as for the quasi-realist account, does not change the issue here. The person must have the power of determination. An ‘immediate agent without’, even if not ‘without’ the cranium, does not answer to this requirement. There must be some measure of ultimate responsibility for there to be any responsibility.

Another way of putting matters is to note that the soft determinists face a squeeze between, on the one hand, the transcendental arguments that can be marshalled whenever we reach a level of decision, ratiocination or choice over which we have fundamentally no control, and, on the other hand, the need to distinguish their various positions from ‘hard libertarianism’.

The self-determinist embraces the latter position: we, ourselves as a whole, have the power to determine our actions – this power is not to be identified with any externally characterizable determination by natural law and will not reduce to the determination of Hobbesian sub-personal mental items; at best ‘internal’ in the sense of being ‘within the cranium’. The action (ratiocination) of a person may ‘taketh beginning from itself’, and not any ‘immediate agent without itself’.

Hard determinism, on the other hand, abandons any claims to freedom, and with this abandons the resources needed to respond to even the simplest version of the transcendental argument – this position becomes self-defeating. If the quasi-realist is really to defang this whole family of arguments, this cannot be merely by surrendering to the transcendental arguments which undermine hard determinism, save only that this surrender takes place when we reach second- or third- or nth-order cognition/conation. Such manoeuvres delay the onset of these arguments, but do not blunt their force – their ability to withhold (here) reflexive epistemic justification. If, however, the quasi-realist never surrenders to these transcendental arguments – if there is a wholly open-ended preparedness to hold that any level of conation/cognition is ‘freely’ chosen – then any such quasi-realism will be unable to distinguish itself from hard libertarianism.

10.3 Mesh accounts more generally

Mesh accounts generally are those compatibilist accounts, whether of free will or moral responsibility, which hold that free will or moral responsibility obtains when there is a mesh between one set of determined things and another – say, first-order desires and second-order desires (Frankfurt 1971) or first-order desires and values (Watson 1975). Although mesh accounts are usually taken to be one, fairly recent sub-species of compatibilist accounts in general, this assumption may be called into question. It is arguable that even classic Hobbesian compatibilism is a mesh account – the limiting case of such: we are free (and/or responsible) just in case there is a mesh between what we are determined to desire (drugs, say) and what we are determined to be able to do (get hold of drugs). The Frankfurt addition to this was to acknowledge that the basic Hobbesian account is inadequate, but to take this insight – hitherto taken to be proprietary to the libertarian – and spin another, higher-level mesh account off it: to hold that the mesh must be between what we are determined to desire (drugs, say) and what we are determined to desire to desire (to be a drug addict – or not). The idea is that free will (sometimes: moral responsibility) is to be found in having a match between one determined set of mental events (say, our first-order desires) and another set of equally determined events (say, our second-order desires).

Although Frankfurt, more than any other philosopher, is associated with the development of mesh accounts, he didn’t rest easy with his initial notion of a mesh. In later work, to the requirement that we have our requisite second-order desires is added the requirement that we identify with said desires.

He may become morally responsible, assuming that he is suitably programmed, by identifying himself with some of his … second-order desires, so that they are not merely desires that he happens to have or find within himself, but desires that he adopts or puts himself behind. In virtue of a person’s identification of himself with one of his own second-order desires, that desire becomes a second-order volition. And the person thereby takes responsibility for the pertinent first- and second-order desires and for the actions to which these desires lead him. (Frankfurt 1988: 53, emphases added)

Frankfurt’s (1975/1988) ‘identification’ epicycle on his original (1971) account looks (despite his denials) to be a further mesh (and a frankly regressive one): between what we are determined to desire to desire and which desires (to desire) we are determined to identify with: a regressive sop to the underlying problem of responsibility that will continue to nag at such accounts, regardless of such additions.

So, some probed this account with thought experiments. What if the agent’s first-order desires were determined (imposed upon that agent – inserted into him, for example, by some neuroscientist/demon) and his higher-order desire also? What if our drug addict were to have had imposed upon him not only his addiction, but also his second-order desire to be a drug addict? Or his higher-order identification with his status as an addict? (These being the various manipulation arguments against compatibilism – e.g. Pereboom 2001; Mele 2006.) Frankfurt responded with further remarks in turn – about the agent’s activity versus passivity (with respect to their identification … with their second-order desires … to have their first-order desires):

What is at stake [regarding ‘acting freely’] is not so much the causal origins of the states of affairs in question [the subject’s first- and second-order desires] but X’s activity or passivity with respect to those states of affairs. (Frankfurt 1988: 54, emphasis added)

As noted in Chapter 7, the problem Frankfurt is here seeking to address is hardly one that is unique to his position, or even the extended family of recent mesh accounts more generally. It is shared by a number of other philosophical positions – existentialist, Nietzschean,³ hard incompatibilist (of various kinds), reflective endorsement accounts (of different kinds) – the need for the final commentary, the final ‘but’ clause, the last word: yes, we’re determined, it’s all futile, eternal recurrence is true or whatever … but … you can adopt an attitude of identification, or higher-order approval, or irony, or defiance, or wryly amused recognition of the absurdity, or stoic resignation, or amor fati, or detachment, or affirmation, or whatever. However, if the pessimistic metaphysics which precedes the ‘but’ clause is true, you can do none of these. An attitude of irony, or defiance, or resignation, or whatever else, is already in the post, to be delivered courtesy of the Big Bang and the laws of physics. Whether there is a ‘mesh’ in determined attitudes or a jagged incongruity isn’t down to you; you have no choice in the matter, even in your most final and pessimistic surrender – or, for that matter, your defiant and brittle irony. I then simply cannot see what entitlement the mesh theorist has to language meant to signal such ownership of his attitudes – that they are a product of his ‘real self’ – and which in the hands of a libertarian would represent acknowledgement of the metaphysical commitments that signalled a terminus to questions of such ownership. Work through this literature and see the requirement that the agent identifies himself with his desires; of these being desires that he adopts or puts himself behind; that the person must take responsibility for his pertinent first- and second-order desires; requirements that the agent reflects critically on a first-order desire before giving ‘higher-order’ approval of that desire … and many, many similar locutions. What means such talk – of ‘higher-order’ approval,4 reflective endorsement and the like? In the hands of a libertarian, I understand well what it means: the claim to a power of arché, the claim to be causa sui, the claim to a power of originative action, of authorship, of terminus authority – in at least some final assent, commentary, gesture or response. But what means this talk in the mouth of the determinist? I can see its rhetorical power, but not the metaphysical entitlement to the role these words are meant to play. So, a much-invoked move was originated by Dennett when he cited Luther’s ‘Here I stand, I can do no other’ (Dennett 1984: 133). Dennett claimed that this may bring home to us that there can be responsibility in a determined world – because we see Luther taking responsibility for a determined act. But that’s surely the point: if determinism is true, not only the act, but the taking responsibility for it, the identification (or otherwise), is determined also. Consider Fischer and Ravizza’s⁵ way of putting matters:

As the child grows up, he is subject to moral education … the child comes to believe that he is a fair target of certain responses – the ‘reactive attitudes’ – and certain practices, such as punishment … it is in virtue of acquiring these views of himself (as a result of his moral education) that the child takes responsibility. More specifically, it is in virtue of acquiring these views that the child takes responsibility for certain kinds of mechanisms. (Fischer and Ravizza 2000: 442; emphases in original)

The child’s ‘taking responsibility’ is, however, itself wholly determined. How can it be an adequate response to the objection that the child doesn’t have responsibility for actions, decisions, ratiocinations, etc., which are wholly determined prior to that child’s existence and outside of his power to alter, to say that the child nevertheless ‘takes responsibility’ (is wholly determined to ‘take responsibility’ prior to that child’s existence and outside of his power to alter) for these actions, decisions, ratiocinations, etc.? Of course, as a point of detail, a second determined process might prove to be required over the first; but arguing that claim would be something that would then be answerable to various scientific, etc., desiderata or minutiae. Here, one doubts that is what is happening. Moving to the second-order level looks like a sop to an incompatibilist objection, one held on stark metaphysical grounds – an objection that is not seen as answerable mechanistically at the first-order level. It is hard, then, to see why it should be satisfactorily answerable in just the same way at the second-order level.⁶ Either flout the incompatibilist objection from the off, or address it fully – this seems to be neither option. That is, one can of course go ‘hard compatibilist’ here; and in some sense, ultimately this must be the option any compatibilist takes: to bite the bullet and stick by one’s preferred compatibilist account, no matter how counterintuitive this be in light of how the poor (e.g. manipulated) victim came by his suitably meshed (etc.) mental states. But note the role compatibilism is (or was) meant to play in the logic of the dialectic that concerns us: to offer the determinist greater retorsive, reflexive, normative–epistemic resources than those offered by hard determinism/hard incompatibilism – in dealing with the transcendental arguments that otherwise threaten determinism as such. How does ‘going firm’ after one iterative cycle (or two, or three) offer the hard compatibilist an alternative, more satisfactory stance than that of the hard determinist/hard incompatibilist who honestly gives the obdurate response from the off?

To need an attenuated, etiolated concept of freedom – merely a ‘last word’ of identification (or pessimism, or irony, or commitment) – is still to need a concept of radical freedom. Thus, to abandon internalist or deontic justification for externalism or consequentialism still requires the final, reflexive ought in putting the externalist or consequentialist theory forward. Everyone, even the hardest, most pessimistic determinist/incompatibilist, must acknowledge some ultimate responsibility in reason.

10.4 Justification, overdetermination, time and the genetic fallacy

Chapter 8 gave a transcendental argument against determinism – the conative transcendental argument – that did not proceed via a deontic leeway principle. The conative argument, however, was not a specifically epistemic argument. It was remarked in Chapter 8 that not even all epistemic transcendental arguments went via the principle of ‘ought’ implies ‘can’ – nor any cognate deontic leeway principle. I wish to give one such argument here: an argument I shall not endorse, but that I believe opens up a line of thought that the determinist nevertheless should find troubling.

A central insight behind any epistemic transcendental argument is that for the determinist to be able to justify her position epistemically, it is necessary that at least some aspects of the adoption, articulation and defence of her position should have been determined by normative epistemic determinants: reasons, judgement, argument, insight, awareness. More plainly: for the determinist to be epistemically justified, her embrace of determinism should have been at least partly determined by epistemic justifiers. I take it that this is self-evident. But the content of determinism itself maintains that its adoption, articulation and defence is wholly, exclusively determined by the Big Bang and a finite set of natural laws, externally conceived. So, this argument goes, determinism cannot possess the reflexive epistemic justification it requires.

The (radical, totalizing) externalist will simply oppose the first and second sentences of the previous paragraph – and claim, on that basis, that all else in that section begs the question. What I will have to say shortly has some relevance to this; but Part 1 of this book has already, exhaustively taken issue with this totalizing, eliminativist position. Aspects of one’s epistemic life may be accounted for externalistically, perhaps large aspects, but not all. And recall: we are concerned with epistemology, not metaphysics here. Further, we are concerned with the epistemology of that epistemology: not just the metaphysics of that epistemology. (The point made in Chapter 5, Section 5.3 to counter Alston and others’ attempted saving distinction – that between ‘being justified’ (the theoretical project) and the ‘activity of justifying a belief, showing it to be justified’ (the regulative project).) Of course, a proposition, belief or argument could be true (or valid), yet its truth/validity not be determined by any justifiers. The claim is that there is one, essential sense of ‘justification’ under which a proposition, belief or argument could not be justified were its adoption, articulation and defence not determined by any justifiers.

This line of reasoning owes something to both the Direct and to the Source arguments for incompatibilism. It is still an epistemic transcendental argument, but one that does not appear to work via any deontic leeway principle. Should Frankfurt arguments be assumed to prevail, or should an ‘asymmetry’ position prove invulnerable to the counterarguments of Chapters 6 and 8, this argument would remain. To repeat: I do not endorse this argument (I fear it may be too powerful, for reasons given below). Nevertheless, I believe it raises large metaphysical challenges for the determinist.

The direct (source) epistemic transcendental argument

1. For the determinist to be epistemically justified, her embrace of determinism (its adoption, articulation and defence) must have been at least partly determined by epistemic justifiers.

2. The content of determinism itself requires that its adoption, articulation and defence was entirely determined by the Big Bang and a finite set of natural laws, externally conceived.

3. The laws of nature were fixed 13.75 billion years ago.

4. The Big Bang was fixed 13.75 billion years ago.

5. 13.75 billion years ago, there were no epistemic determinants of anything anywhere in the universe.

6. So, the determinist cannot be epistemically justified in her embrace (adoption, articulation and defence) of determinism.

In response to this argument, the most likely move for the determinist is to claim that her embrace of determinism is in some way both determined by the Big Bang and by normative epistemic determinants. Focus not on the details of such accounts, but at a more purely metaphysical level, ask how this can be: how can the normative epistemic properties that, reflexively, the determinist requires to have been the determinants of her advancing and inhabiting her position then coexist with the exclusively causal determination of her advancing and inhabiting her position? For example, which of these has priority? (Temporally, it is entirely clear which has priority.) Note that we are not here concerned with the specific details of some single, proprietary, epistemically externalist account of justification (Goldman, Nozick, Dretske, etc.), nor yet with similar, specific, fine-grained details of this or that compatibilist account of free will (Hobbes, Frankfurt, Fischer, etc.). We are concerned with the bigger view of co-determination that may be seen as underpinning these.

The bigger metaphysical picture

In broad outline, the range of responses available to the determinist here will closely resemble those found in the structurally similar special science and mental causation literature as this developed from Descartes onwards – and in the previous chapter (after acknowledged reservations regarding the exclusion problem) I guardedly mentioned my own preferred option in that literature (emergentism, of a self: with attendant determination by that self in interaction with its reasons). Notice, though, that when the various positive responses to overdetermination have been advanced by anti-reductive philosophers of mind, science and metaphysics, they have been vigorously attacked by a large subset of those determinists who now have need of them in turn (see Endnote 9 at the end of this paragraph). How can normative epistemic properties determine adoption of the belief in determinism and this belief have been entirely determined by the Big Bang and the fundamental laws of nature? The problems which attend attempted solutions to the structurally similar problem in the philosophy of mind and special sciences include among them at least those of escaping epiphenomenalism (Davidson 1980),7 problems of empty, terminological evasion of the problem (Smart 1959)⁸ and the ‘Exclusion Problem’ (Kim 1989).⁹

Severe as they are when they appear in the philosophy of mind and special science literature, these and cognate problems become that much more severe when we factor in the temporal component intrinsic to determinism. To contrast our case with a stock special science example: it may already be difficult to make a river’s determining the erosion of its outer bank something that is metaphysically compatible with the quark-level determination that apparently competes for this role; but at least each of these putative determinants (river, quarks) operates at more or less the same time.¹⁰ Consider, though, our case.

The content of determinism requires that the defence of this position (the arguments for, convictions about, articulation of and actions rationally consequent upon determinism) should all of them have been entirely determined by the Big Bang, thereby entirely fixed and immutable from the moment of that singularity, 13.75 billion years ago. But at that time, there were no normative epistemic properties determining anything at all: no reasons, etc., would determine anything anywhere for about 13.74 billion years. Those normative epistemic reasons (arguments, evidence), on the basis of which the determinist must claim to embrace and defend her position, are, according to the content of her own theory, things that, at the time that her theory’s adoption was determined, could not have been determinants of anything in the universe. So determinism must be without epistemic justification.

Objection: ‘The Direct Argument is parasitic on leeway incompatibilism’

One influential counter to the Direct Argument’s use of transfer principles has been addressed already in Chapter 6 (Section 6.6.3; Ravizza and others’ Frankfurt-style pre-emptive overdetermination arguments – such as ‘Erosion’). After Widerker (2002), a distinct challenge is that these arguments tacitly contain the principle of alternative possibilities/‘ought’ implies ‘can’ as a suppressed (framework) premise. This challenge is that the premise that there can be no (here) epistemic responsibility for events occurring 13.75 billion years ago invites the question: ‘why not?’ And the answer, this ‘parasitic’ criticism maintains, must be: because no-one could alter these things.

It is questionable, however, whether this is a good objection to seeing this family of incompatibilist argumentation as being distinct from leeway incompatibilism. Haji’s (2008) response is that ‘fixity’ premises of this kind may be vindicated simply because the states of affairs they describe (laws, Big Bang) obtain independently of the existence of anyone – or, stated causally, that no-one is causally responsible for the existence of such states of affairs.11 McKenna (2008) notes that the fact that the Big Bang and the laws of nature were temporally prior to anyone existing at all – and further the fact that no-one can produce a law of nature (or could produce a singularity) – seems relevant.¹²

A profound impossibility is enough to ground a philosophical argument; and as noted in Chapter 8, it is a philosophical argument that is being advanced here, not a ‘proof’. On the determinist’s own views: (a) at the time of the Big Bang, no epistemic properties could have determined the determinist’s embrace of determinism; and (b) the determinist’s embrace of determinism was entirely determined at the time of the Big Bang. The determinist must maintain that all aspects of the defence and articulation of determinism were already and entirely determined billions of years prior to the existence of any epistemic determinants of anything in the universe – that is her thesis; it is not something imposed upon her.

Whether or not this response is sufficient to overturn Widerker’s ‘parasitic’ objection, as a ‘wide source incompatibilist’ (one who wishes to defend leeway as well as source incompatibilism), I would be unfazed if in some deep sense a tacit reliance on leeway turned out to be contained somewhere within the direct/source transcendental argument. Conceptual connections between source, leeway and ultimacy arguments for incompatibilism are undoubtedly present, albeit difficult to draw out; but these connections appear to go in more than one direction of dependency at once. I should be content were each such core motivation found to be as fundamental as each other.

The genetic fallacy objection

A distinct, earlier response to the direct (source) transcendental argument is the contention that it embodies a genetic fallacy (e.g. Anscombe 1981: 226). The stock realist dismissal of the genetic fallacy is to distinguish a ‘logic of discovery’ (here: determination by the Big Bang) from a ‘logic of confirmation’ (here: current best arguments for determinism). The objection will be that it matters not how we come by our belief in determinism, it matters only that the belief is true, or the process by which we arrive at it is truth conducive (Anscombe herself uses the example of a reasoning process being valid).

In this case, however, any such response does not appear to be an easy one to make good. Firstly, bear in mind the point emphasized especially in Chapters 5 and 7: we are here concerned with epistemology (that is, in this avowedly normative epistemic context, there is a limit to how ‘realist’ we can go in regard to drawing an epistemology/metaphysics distinction). Then realize that we are concerned with determinism as a totalizing thesis. The legitimate deployment of this realist distinction in other, more local contexts takes place against a backdrop of there being avowedly some normative epistemic properties: those resources with which we evaluate the quality of the logic of confirmation (those normative epistemic resources with which we decide whether the argument, however it came to pass, is a good one). These resources, though (to evaluate the logic of confirmation), were, according to determinism, themselves entirely determined many billions of years before any normative epistemic properties were acting to determine anything.

So, when Kekulé discovered the ring shape of the benzene molecule upon waking from a reverie of a snake seizing its own tail, the origin of his position in a non-rational process (his dream) did not impugn his discovery because the outcome product of this process of discovery (his published hypothesis about benzene’s shape) is subject to its own normative epistemic checking procedures – those of the organic chemists in his scientific community. They didn’t check Kekulé’s discovery by dreaming about it themselves, in turn. We (if realists) waive any right to criticize or subject to normative epistemic scrutiny the genesis of an idea, argument or theory, because, by hypothesis, we have sufficient freedom from non-rational (causally-genetic) determinations at the point of outcome: in assessing this idea, argument or theory for truth (or validity, or evidence, etc.). However, this state of affairs stands in marked contrast to the case in point – that of determinism. Both the origin of the hypothesis (‘discovery’) and its testing ground in argument, reasons, etc. (‘confirmation’), are envisaged as having been as determined as each other, which is to say wholly determined, by just the same singularity, 13.74 billion years before normative epistemic determinants were determining anything on earth.¹³

This points up the fact that, in contrast to such piecemeal and workaday examples as Kekulé’s dream, there are complete metaphysical positions which, given their content, are vulnerable to being reflexively undermined due to the all-encompassing nature of their own claims about their determined origins. That is, there are ‘totalizing’ metaphysical positions for which the distinction between the logic of discovery and the logic of confirmation is unavailable. The most notorious of these is relativism – it exemplifies most of what is wrong with the rest.14 To claim a complete social determination of all human thought requires that this view (relativism) in turn should have been socially determined – and the peritrope problems for that position, familiar to us from the Theaetetus onwards, emerge. These problems are unsolvable by this and similar positions. Making such peritrope arguments into problems for (here: social) determinism is not a question-begging contrivance of the opponent of such positions, but something that develops out of the content of the target position (determinism, relativism, eliminativism) itself. I have defended this view for these other cases elsewhere and at length (Lockie 2003b, c, 2006).

Objection: The direct transcendental argument is much too powerful (the abiogenesis objection)

The strongest objection, an objection I shall not accede to, but which is strong enough to stop me from simply, wholeheartedly endorsing this direct (source) transcendental argument, is just that it is too powerful – that it must go wrong somewhere. This, however, is a wholly negative counterargument; were it to be conceded, it would nonetheless leave undiagnosed where the Direct Argument goes wrong, and thus acknowledging the force of this counterargument would continue to leave work for the determinist to do: it would continue to be a challenge for determinists to state how they would resist this argument (see remarks under ‘the bigger metaphysical picture’ above). But a case can be made that it is too powerful.

Life emerged perhaps 3.7 billion years ago, and there were no biological properties 13.75 billion years ago. Yet the conjunction of physical determinism with the thesis that there are true biological determinants of occurrences over the last 3.7 billion years is not obviously problematic – or so this objection goes. If true biological determination can emerge out of a physically deterministic system, then why cannot true epistemic determination emerge likewise?

Well, a caution here: there is no analogical extension of the above argument to commit its defender to the view that true biological determination cannot emerge out of a physicalistically deterministic system through abiogenesis – in fact, quite the contrary. Rather, it is plausibly an analogical extension of the above argument that after true biological determination indeed unproblematically emerges out of a physicalistically deterministic system, the era of true physical determinism must then be seen as having come to an end (physical determinism, mind, not physical determination). A structural analogue of this argument would plausibly require us to argue that modern biologists can unproblematically retain abiogenesis, but that they may do so only by insisting that after life’s emergence, the presence of biological determinations in the universe prevents physical determinism (a totalizing thesis) from continuing to be true. That is what a commitment to true biological determination and Alexander’s Principle (or something like it) requires. But is this entailment of the argument itself quite damaging enough?

Well, maintaining the latter line is not by any means absurd. As indicated, reductionism may make things difficult for many special sciences, including the biological sciences (the classical emergentists accordingly had ‘life’ as one of their major emergent strata). The other great emergent strata 15 (chemical, biological, perhaps non-agential psychological) may indeed conflict with physical determinism – the special science literature may be seen precisely as identifying that conflict.¹⁶ It may also be noted that the properties of life – homeostasis, replication, metabolism, etc. – do not obviously include anything resembling full-scale agency, and that it is agency above all that makes things difficult for the determinist. Agency, if it is anything, is a power to determine action. And agency fulfils a normative role (cf. Chapter 7, Section 7.2.2). Accordingly, agency appears to implacably conflict with (compete with) any thesis as to physical determinism’s exhaustive prior determination of action.

Whether this argument is over-powerful, then, is controversial; but over-powerful or not, one should expect a bigger metaphysical picture from the determinists in response to the reflexive problems such arguments present regarding source, ultimacy and the emergence of the normative from the non-normative in a world exhaustively fixed prior to the existence of the normative.

10.5 Opacity and presupposition

The claim of the previous chapter was that the freedom needed for epistemic (and other) responsibility is a freedom of self-determinism. This position may now be related to the material considered in Chapter 8 (Section 8.3.2) on transparency versus opacity in argumentation.

From Chapter 8, we saw that for the compatibilist, ‘able’, ‘unable’, ‘avoid’ and ‘alter’ – all the modal verbs of intentional achievement, etc. – are trans-substitutionally opaque, analogous to the propositional attitudes in the philosophy of language. For the incompatibilist, able, unable, etc., are transparent. That is, for the incompatibilist, of course if you are powerless to change the laws and start-up conditions – given that these would need to be different to perform an action p – then you are powerless to bring about p.

For the compatibilist, just as Lois Lane can want to kiss Superman but not want to kiss Clark Kent, so I can have the ability to perform an action p, yet not have the ability to change the laws or conditions that would in fact need to be changed to permit this to happen.

Are these two possibilities – transparency and opacity – the only ones available to us? Consider again the neo-Moorean argument given in Chapter 8. Some philosophers (Wright 2002; Davies 2004) oppose this argument by claiming that warrant doesn’t transfer from M1 (‘Here is one hand’) and M2 (‘If this is a hand, then I am not a brain in a vat’) through to M3 (‘I am not a brain in a vat’). Others (Beebee 2001) claim that M1–M2 may well offer warrant for M3 as far as it goes, but that statements like M3 are tacit, framework presuppositions required for the warranted utterance of statements like M1. Since M3 is presupposed by M1, M1 can hardly then be support for M3 – on pain of begging the question.

In either such case, it is held that we can’t use Moore’s M1 and M2 to constructively argue for M3. But if we take something like the Beebee line, we can say M3 is presupposed in the proposition asserted in M1 – indeed, in pretty much any proposition, p, one cares to assert. Denying M3 is not then in any obvious sense an option. M3 is indispensable – though not ‘warranted’ from the warrant possessed by statements like M1 in any conventional, constructive (say, evidence-based) sense of ‘warrant’. M3 is not, however, in any sense thereby unwarranted – indeed, it is a presupposition of other things (any other things, like beliefs about hands) having warrant. M3 must be presupposed yet cannot be proved (at least, in any sense of ‘proof’ continuous with the sense in which M1 is provable); or so a version of this argument would go.

Now consider those who object to the transfer principle in the consequence argument:

2. Necessarily, given a complete state of the world at some time in the remote past and all the laws of nature, then p, for any arbitrary p.

3. We are powerless to avoid or alter the remote past or the laws of nature.

4. We are powerless to avoid or alter p, for any arbitrary p.

The libertarian, along with the soft determinist, takes as a given that we can do something that experience, ordinary language, common sense, pragmatics, etc., would hold we can do – a ‘p’, any arbitrary p (raise one’s hand). This is like Moore’s ‘here is one hand’. But unlike the compatibilist, the conventional libertarian does make a neo-Moorean move: employing closure to argue that determinism (or, for a more sophisticated version of the argument, the determinism/indeterminism dilemma) is false (or a false dilemma). In the older, now somewhat unfashionable tradition, the libertarian did this as a result of the data of his own experience as of freedom. Again there are parallels with Moore – just as Johnson refuted Berkeley ‘thus!’ and kicked his stone, so he held ‘of course the will is free and there’s an end ont’!’ And just as for Moore’s argument, with transparency accepted, one philosopher’s modus ponens may be another’s modus tollens: the hard determinist will argue from 2 and 3 to 4.

In contrast to these ‘transparent’ incompatibilists, the compatibilist denies that we can transfer warrant from our motivated opposition to 4 (namely, the warrant we have to think we are not powerless to avoid or alter p, for any arbitrary p) to conclude by modus tollens the refutation of 2. No less than this, the compatibilist denies that we can transfer warrant from our motivated endorsement of 2 to conclude by modus ponens that we are powerless as in 4.

These argumentative contexts are opaque, says the compatibilist. The warrant we have for 4 is (say, with Hobbes) that we can ‘do as we will’. The warrant for 2 is our best theory of physics – or was once, when physics was assumed to be deterministic. The point is, deterministic or indeterministic, the standpoint of physics is external. The objection to the transparency incompatibilists made by the opacity compatibilists is to deny that we can argue from Campbell’s ‘inner standpoint’ to a substantive claim about his ‘standpoint of external observation’17; that is, to deny that my experience as of freedom, embedded within a context of justification answerable to the inner standpoint, can urge a substantive claim about physics, embedded within a context of justification answerable to the standpoint of external observation.

Admittedly, the reverse kind of parochialism – of enthroning one standpoint above the other – is far more common: enthroning the theoretical standpoint, the standpoint of external observation, over the inner, or ‘practical’ standpoint. This direction of parochialism is now more common even than when C. A. Campbell wrote:

if determinists could provide convincing theoretical arguments against a free will of this [incompatibilist self-determinist] kind, the awkward predicament would ensue that man has to deny as a theoretical being what he has to assert as a practical being. Now I think that the determinist ought to be a good deal more worried about this than he usually is. He seems to imagine that a strong case on general theoretical grounds is enough to prove that the ‘practical’ belief in free will, even if inescapable for us as practical beings, is mere illusion. But in fact it proves nothing of the sort. There is no reason whatever why a belief that we find ourselves obliged to hold qua practical beings should be required to give way before a belief which we find ourselves obliged to hold qua theoretical beings; or, for that matter, vice versa. All that the theoretical arguments of Determinism can prove … is that there is a radical conflict between the theoretical and the practical sides of man’s nature, an antinomy at the very heart of the self. And this is a state of affairs with which no-one can easily rest satisfied. (Campbell 1957: 158–79)

Campbell has, to my mind rightly, identified two things here that other, lesser philosophers often hold separately but not conjointly. He holds with the ‘transparent’ incompatibilists that propositions like the denial of 4, vouchsafed to us by his inner standpoint (the practical realm), do conflict with propositions like 2, vouchsafed to us by his standpoint of external observation (the theoretical realm). But he also holds with the ‘opaque’ compatibilists that these propositions do indeed embed within radically different standpoints (perspectives, schemata, Gestalten). I do not see how we can deny either of these claims – except on the most unprincipled grounds: of effacing one or other apparent truth merely for the sake of making our position defensively less exposed.

Those philosophers who are convinced, on other grounds, of the correctness of some one particular, proprietary, compatibilist analysis (say, the Hobbesian, or one of its successors) will have reason to carry that ingenious formulation forward like a shield and just stand fast upon the claim of opacity; but no-one else will have reason to accept the claims of opacity and then simply, irenically, rest with these claims. Specifically, the separately warranted beliefs that you have freedom (not 4) and that determinism is true (2) absolutely will not, in and of themselves, give you any reason to believe in opacity. These twin beliefs will only give you reason to agonize over the fact that the consequence argument shows that two positions you wish, on distinct grounds, separately to believe in appear nevertheless to be incompatible.

Equally, however, it seems wretchedly unsatisfactory to argue from the utter indispensability of freedom – which datum is embedded within the practical, inner standpoint – to some positive view (we yet know not what) about, what, indeterminism in physics? Neither indeterminism nor determinism in physics? Some novel position on the conservation of mass–energy? It seems scarcely less unsatisfactory to argue by contraposition from the dilemma of indeterminism/determinism to some ‘hard’ successor view of incompatibilist non-libertarianism. The practical/theoretical Gestalten are just radically unalike: the schemata their propositions embed within do not permit easy alignment with each other, and we must be cautious about drawing any swift conclusions in such cases. It is unsurprising that some regard this as more than a philosophical problem, but as an unsolvable mystery (Chomsky 1988: 151–2, 2004: 382–3, 2016: 53–4 and Ch. 4; McGinn 1993; Kim 1989 – e.g. 96–7; Nagel 1986; van Inwagen 2000; Kant 1933: 3rd Antinomy).

If we cannot solve a McGinn/van Inwagen mystery, it is surely at least incumbent on us to try to explain why we can’t solve such – why it is a mystery. These two standpoints represent very major, intertwined groupings of schemata: practical versus theoretical. Propositions that are intelligible only inasmuch as they embed within the one grouping of schemata are then compared with propositions that embed only within the other grouping of schemata: mystery results. The standpoint of external observation answers to a different Gestalt to the inner standpoint – of practical freedom.18

This, as far as it goes, starts to resemble the position of certain other ‘two-standpoint’ philosophers: that we note the different (opaque) contexts represented by the practical and the theoretical standpoint. But all too often is added the rider that we should stop worrying about conflict since each of our premises belongs to a different sphere (whilst at the same time pretty clearly according the theoretical sphere a metaphysical or other priority).¹⁹ I think that this, as a positive view, is far too irenic, too quietist, too complacent. It is true that the internal and the external answer to different constraints, but we can’t help drawing the alethic consequences and we aren’t wrong to draw these consequences. We see the conflict – it hurts. Other ‘two-standpoint’ positions have tended to be compatibilist – mine is not. But I do not think we can go to a conventional, constructive, ‘modus tollens’ libertarian incompatibilism either.

I want to suggest, at least at this stage in intellectual history, that we be presuppositional (Kant’s, though not my ‘regulative’) rather than ‘constructive’ (Kant: ‘constitutive’) incompatibilists. We have to embrace freedom, and we note that this presupposes an ability to do otherwise than determinism would have it. This isn’t an optional standpoint, one that is ‘useful’, say, for us to choose to adopt (that we might equally set aside). But we can hardly claim thereby to have a constitutive proof of determinism’s refutation. Determinism’s acceptance or positive refutation answers to a different Gestalt (standpoint, context, perspective) than that of freedom: theoretical rather than practical, external rather than inner. It’s not that we can then, irenically, leave it at that; this is not an acceptable state of affairs – it’s not as if we can say, ‘So there’s no conflict then,’ and carry on without metaphysical distress. Just as for the Moorean epistemic example, there is conflict; but the separate premises do answer to different schemata. We must presuppose freedom within the practical Gestalt. We must then address the forced choice between determinism and indeterminism within the almost incommensurable theory structures of natural science (sometimes physics, sometimes other sciences, with no surety that these physical, psychological, etc., schemata are themselves compatible). We note the conflict between the thesis of freedom and that of determinism – and indeed physical indeterminism. We note also that the content of these theses – of freedom, of determinism/indeterminism – themselves are embedded within separate schemata; and so we cannot even be certain as to the degree of implacability of this conflict.

This position is that of presuppositional incompatibilism. Abandoning freedom, radical freedom, is not an option; and an ersatz, reductive construction thereof does not avail us. But our presuppositions within the practical sphere cannot claim the status of positive, constitutive results within the theoretical, external standpoint – for theory construction within this standpoint answers to, and embeds within, quite other schemata. This is transcendental presupposition. We are practically free – free within the domain of appearances (all possible appearances) – with a failure to appreciate this point leading to unfortunate passages like that offered by Grünbaum in Chapter 8.²⁰ Moreover, all appearances transcendentally presuppose more – though they cannot constructively prove more.

Epistemology aims at truth (Chapter 5): it is intrinsically truth directed. And truth is transparent: we can, philosophically, follow the arguments, and see the trouble we are in. Even in recognizing the fact that practical freedom and theoretical determinism/indeterminism embed within different schemata, we can’t rest with this – any more in the conative than in the epistemic case. Moore’s epistemic argument shows us something is profoundly wrong with regard either to our presumed knowledge of hands or our presumed possibility of philosophical doubt. And the consequence argument does likewise with regard to our necessary presupposition of freedom in the teeth of indeterminism and determinism alike.

We can’t bracket a clear alethic consequence on grounds that an epistemic practice or an ordinary language usage or a proprietary formula has it that transfer fails. I can see the validity of the Moore argument – so can you. It has no epistemic operators in it – we are (rightly) reading these into it afterwards, because that is its philosophical context; but alethically it is valid, and we can (and must) see that, too. In just such a way there is still a mystery surrounding free will and determinism. But recognizing something of why a mystery arises is progress, too, of a sort.

10.6 The issue of transcendental realism versus idealism

Transcendental or Epicurean or peritrope arguments are negative arguments – they apply against positions. A transcendental argument is a type of intellectual jujitsu. Its characteristics are as follows:

1. It must always be against an extant or possible position.

2. This target position must be a deflationary position.

3. This target must be a sweeping, overarching, ‘totalizing’ position.

In responding to a transcendental argument, there will then be either of two basic escape strategies available: to make the target of the transcendental argument an irreflexive position – one that doesn’t apply to itself – or to make it less deflationary than it appears at first. We saw that the latter position – trying to make determinism softer and less deflationary – ends up with determinism regressively chased up a ladder of transcendental arguments until it faces a forced choice between hard libertarianism and hard, futile determinism. The former position, though – that we should make determinism irreflexive – is hopeless. Even if one only has enough freedom to take epistemic responsibility for the argument as to one’s lack of freedom, the property of freedom comes into the world. For there to be even just a ‘final ought’, a ‘last word’, there has to be a space for reason, and the freedom needed to inhabit that space is ineliminable.

This engages more generally with the eliminativist options of embracing an ‘existential’ (totalizing) rejection of deontology (for the ethical transcendental argument) or internalism (for the epistemic transcendental argument). Consider again arguments like Grünbaum’s in Chapter 8. Of course we can argue that this or that badly damaged moral agent had only an attenuated degree of freedom, or indeed no freedom, and thus must not be judged harshly – or perhaps at all. Of course such agents may not be fully responsible for their actions, and may at best be subject to a consequentialist ethics. But what of the situation of we who are reaching this conclusion? Or being urged by argument to reach this conclusion?

As for ethics, so for epistemology: one may point to the limits to any Cartesian picture of our mental powers, note our lack of ‘cognitive voluntarism’ and recommend a strongly externalist epistemology. But what of we, the epistemologists, cognitive scientists and the like, who are arguing for that claim? The last ‘ought’ – that we ought to embrace externalism, that we would be irrational not to – is ineliminable and, in a certain crucial sense, prior. To argue for a strong externalism, one always needs the ‘last word’, the last ought. It is easy to argue for a (very strong) externalism as long as one forgets this. The determinist must be arguing for a totalizing externalism. In doing this, the externalist will end up as the ‘subject who fails to take account of himself’ (Schopenhauer 1966: 13). The last ought is ineliminable, and ought implies can.

Those who resist these transcendental arguments have sometimes approached the points being expressed here without realizing how much they are conceding. So, Anscombe responded thus to C. S. Lewis’s attempts to develop a transcendental argument against what he called ‘naturalism’ and we would call determinism:

[Of the ‘naturalistic hypothesis’] … you say ‘But if this were so it would destroy the distinction between valid and invalid reasoning’. But how? … What can you mean by ‘valid’ beyond what would be indicated by the explanation you would give for distinguishing between valid and invalid, and what in the naturalistic hypothesis prevents that explanation from being given and from meaning what it does? (Anscombe 1981: 226, emphasis in original)21

What the determinist has needed all along to escape the transcendental arguments is this autonomy of reason; a normative epistemic and ethical autonomy from Anscombe and Lewis’s ‘naturalism’ – the domain of natural law. We saw earlier that the determinist’s right to such an autonomy may be queried – such a passage raises the spectre of overdetermination problems – but grant it. The aim of these chapters was to develop a transcendental argument for freedom, not necessarily against determinism. A radical, necessary independence, an autonomy of purely internally characterizable reason, must be conceded, and not as an optional concession, but as all that stands between determinism and incoherence. This autonomy of reason is not, as it were, at the level of empirical hypothesis or constructive, constitutive analysis. It must be presupposed, as something prior to doing our naturalistic empirical sciences – our physics, our cognitive science – or our philosophy.

This idea of presupposition raises serious questions as to whether these transcendental arguments, and this freedom they defend, is something we can claim to be transcendentally real (Kant: per rationes essendi); or whether instead the preconditions for this freedom are wholly beyond our ken, and all we can conclude from these arguments is that we have to act under the supposition of such freedom, making its presupposition a (Kantian) regulative condition of rational thought, but not an object of knowledge (Kant: per rationes cognoscendi). I think a good deal remains to be said in the constructive articulation of the self-determinist viewpoint as empirical psychology – a psychology of emergentist, holistic, anti-reductive agency – in which the self and its reasons are irreducible to associative non-rational components: irreducible to Hobbes’s ‘action of some other immediate agent without’. Nevertheless, whilst prepared to defend the truth of such a picture as (perhaps very high-level) empirical psychology, I wonder whether it will be adequate to address the more severe, more purely metaphysical problems of free will and determinism: problems of overdetermination, of exclusion, of conservation, of ‘downwards causation’. Perhaps it will be adequate to this task; but were this positive programme unable, ultimately, to solve the biggest metaphysical problems, still the negative point deriving from the transcendental arguments would remain: we have to presuppose freedom.

To say freedom is something that is transcendentally presupposed is not to say, with the great family of reductive compatibilisms, that such a freedom is not radically incompatibilist, is really a construction, a variant of the various formulations of ‘doing as we can’, etc. Regardless of the specific content of the views they develop, the activities of the physicists and philosophers in their exercise of reason, alike transcendentally presuppose freedom. Yet the free exercise of reason in doing, say, physics, one of its loveliest products, can yield a belief in a purely externally characterizable determinism or (positive) indeterminism, either of which appears complete in itself, and incompatible with freedom. Can it be made intelligible how the freedom we must rationally presuppose could coexist with the universe physics tells us there is – whether determined or not? This is a very tough question, one to which Kant, in the end, replied in the negative:

Nothing is left but defence – that is, to repel the objections of those who profess to have seen more deeply into the essence of things and on this ground audaciously declare freedom to be impossible. (Kant 1964: PA 459)

Can it be made intelligible how an anti-associationist empirical psychology of emergent self-determined holism could be sufficient for the freedom that this chapter’s, and this book’s, transcendental arguments compel us to assume? Perhaps: the positive account given in this work travels some distance in that direction (and certainly I take it to be true as far as it goes), but, with Kant, I doubt whether it will go far enough. Can we, however, make intelligible how the requirements of freedom might be a kind of rationally necessary illusion? To the extent they were necessary, they would not be an illusion.