Incompatibilism and the Principle of Sufficient Reason in Kant’s Nova Dilucidatio

2.1. Creation and freedom

In the Critique of Practical Reason, Kant famously dismisses Leibniz’s conception of freedom as ‘nothing better than the freedom of a turnspit,’ a contraption that in a sense generates effects ‘once it is wound up’; analogously, Leibnizian monads are wound up at creation like bits of spiritual ‘machinery’ (5:97; cf. 28:267). Although a monad is the immediate cause of its own states, it is just as passive as a turnspit with respect to the nature with which it was created. Once this nature is determined, on Kant’s reading, the complete history of the monad is fixed.

Additionally, Leibniz’s system raises the suspicion that God turns out to be the author of human sin, even though Leibniz deployed various strategies to make room for human responsibility, such as a distinction between absolute and hypothetical necessity. One philosopher who pressed this worry was Christian Crusius. He was particularly worried about the role of Leibniz’s version of the principle of sufficient reason.⁴ As we’ll see, Kant engages with Crusius on this topic in detail in the ND.

The standard reading of Kant’s early career, summarized by Christopher Insole, is that up until around 1763 he remained ‘content’ with a compatibilism ‘whereby we are determined by the chain of determining grounds that rests ultimately in God’ (2013: 67). If the turnspit worry occurred to Kant at all during this period, on this reading, it did not trouble him (Insole 2013: 74).

Little attention has been paid to the fact that already in the ND, Kant raises versions of both the turnspit and authorship worries about Leibniz’s conception of freedom, and takes these worries to be interconnected. Additionally, he denies that Leibniz can resolve these problems with the absolute–hypothetical necessity distinction. Regarding the worry that God is the author of sin, Kant writes:

If, as happens in the case of machines, intelligent beings were able to comport themselves passively in relation to those things which impel them towards certain determinations and changes, I should not deny that the blame for all things could be shifted to God as the Architect of the machine. But those things which happen through the will of beings endowed with understanding and the spontaneous power itself of self-determination obviously issue from an inner principle, from conscious desires and from a choice of one of the alternatives according to the freedom of the power of choice. (ND 1:404)

Here Kant connects a worry about God’s authorship of sin to the idea that being the source of one’s actions does not rule out being a merely passive ‘machine.’ That is, if finite agents were merely causally passive in the face of a prior series of grounds, ‘blame’ really would be shifted to the ‘Architect of the machine.’ This indicates that Kant’s references to ‘inner’ causal powers in the second half of the quotation are meant to contrast not just with the causal powers of other created beings, but with God as creating the world and all of the agents in it. It is noteworthy that in the mid-1760s, when Kant is widely agreed to endorse a libertarian conception of freedom, he uses the same language. Freedom is contrasted with being ‘passive’ in relation to ‘a highest productive cause’ such as God, as well as with respect to created ‘objects’ that ‘affect’ us (Refl 3856 17:314).

Returning to the 1755 text, Kant goes on to link moral imputability to acting on an ‘inner principle of self-determination’ (ND 1:405). In context, the relevant principles would need to be independent of God’s choice to create agents with particular natures: therefore, it would apparently need to rest on something like a power to do otherwise.

He also asks why an omnipotent God ‘permitted’ sin rather than preventing it, and provides an answer that reveals his thinking about freedom (ND 1:404). Permitting sin is not the same as causing it, and Kant contends that God can only prevent sin by ‘warnings, threats, encouragements, and furnishing the means,’ because only these indirect divine actions could be ‘compatible with human freedom’ (ND 1:404–5). These limitations on divine action suggest a conception of human freedom that allows for alternative possibilities. God permits sin because the only alternative would involve suppressing human freedom. Contrast a Leibnizian account. If God creates agents such that they necessarily will sin, these agents would be free, for a Leibnizian, because they act with source spontaneity.

This reading is reinforced by notes dating from a year or two before the publication of the ND. Kant is considering questions of theodicy: specifically, why a perfectly good God would allow for moral evils.

Freely acting beings might have avoided many evil actions, and they would have greatly pleased God had they done so. However, the choice of the lesser of two evils, of which one was the lack of freedom and the other of the morally best, was an unavoidable necessity. … Gratitude is due to the Eternal Wisdom for having admitted only the smallest amount of evil. (Refl. 3304 17:232)

Note the counterfactual language in the first sentence, which suggests an incompatibilist account of freedom. In actuality, freely acting beings have committed many evil actions, but they might have done otherwise. As the next sentences make clear, Kant’s view is that God allows freely committed moral evil because it is a better alternative than eliminating creaturely freedom in order to make beings that always acted morally perfectly. Absolutely speaking, God could prevent moral evil, but chooses not to do so because this would impinge on creaturely freedom. There is no suggestion that hypothetical morally perfect beings would lack the source spontaneity required by a Leibnizian account of freedom. Since creatures’ freedom would ex hypothesi be undermined in this scenario, freedom must consist in something more than mere source spontaneity.

Kant further distances himself from Leibnizian compatibilism by denying that the absolute or logical contingency of our actions can help here, so long as these actions are still taken to be hypothetically necessary.⁵ The fact that it was in God’s absolute power to have chosen to create a different world is cold comfort for us if it is still really impossible, given the world that God actually created, for us to do otherwise (ND 1:399). For then, it is not in our power to have done otherwise, holding the rest of the created world fixed. With these criticisms, Kant may well be observing a standard medieval distinction between real modality and what was often called logical modality. This distinction was taken as a way of specifying more precisely what freedom of decision requires, rather than as ruling it out.⁶

Kant’s rejection of these Leibnizian strategies has been noted by Byrd (2008: 73), who concludes, however, that Kant endorses necessitarianism. Necessitarianism indeed threatens if one (i) reads Kant as a Leibnizian compatibilist about divine and creaturely action; (ii) takes seriously his rejection of Leibnizian ways of avoiding a slide into necessitarianism; and (iii) observes that he does not articulate alternative strategies for avoiding necessitarianism.

I favor avoiding a necessitarian reading by denying (i). But some passages in the ND do lend apparent support to Byrd’s necessitarian reading, and need to be discussed. Kant holds that divine actions are morally necessitated. He goes on to state that moral necessity has no less ‘force or … effective power’ than absolute necessity; therefore, in particular, ‘the act of creation is not indeterminate in God, but … is so certainly determinate that the opposite would be unworthy of God’ (ND 1:400).

A look at the context of this striking passage, however, indicates the crucial role of semantic and epistemological assumptions about the certainty of a truth or proposition (ND 1:400). The key doctrines Kant wants to defend here are the ‘certainty’ of divine foreknowledge, as well as our ‘certainty’ about God’s choice to create (1:400). The ‘force’ or ‘power’ of moral necessity can be read as concerning the certainty, for some agents, of truths or propositions.

It might be objected that if God is certain about a truth, then that truth will be necessary. But the assumption made behind this objection cannot be taken for granted in Kant’s historical context. It was routinely denied by late Scholastics in the context of debates over divine foreknowledge, and was even denied by Leibniz in his bid to distinguish his position on creation from necessitarianism (Murray 2004: 18–19, 27 n. 53). That is, it was common to think that the certainty of divine foreknowledge does not entail the necessity of the objects of divine foreknowledge. This would be true a fortiori for whatever certainties finite beings like us are capable of. Kant’s own view is that creatures’ free actions are ‘determined by their own grounds’ and not by divine foreknowledge. If foreknowledge necessitated our actions then source spontaneity, as well as indeterministic spontaneity, would be ruled out (ND 1:405). So Kant’s remarks do not entail that on his view, moral necessity collapses into absolute metaphysical necessity.

Note too that this passage does not literally claim that failing to create is beyond God’s power. The claim it makes is normative: such an act would be unworthy of God’s perfect goodness. This sort of merely deontic interpretation of moral necessity can be found in Samuel Clarke, and sometimes in Leibniz. On this understanding, as Robert Adams has put it, ‘the morally necessary is what one morally ought to do,’ and does not eliminate the possibility of acting otherwise (Adams 1994: 22). The late Scholastics had developed more metaphysically loaded conceptions of moral necessity. But these too, as Michael Murray (1995; 2004) has discussed, were meant to avoid standard causal and metaphysical determination, though without making action brute or unintelligible.⁷ In sum, while moral necessity was conceived in a number of ways in Kant’s time, and it is debatable just how Kant understood it, a point of consensus was that moral necessity does not amount to strict or absolute necessity.⁸

A wider look at Kant’s writings in the 1750s provides further evidence against a necessitarian reading. Kant’s definition of the principle of sufficient reason as ranging only over contingent properties is incompatible with necessitarianism (ND 1:396). Kant’s central argument for this principle relies on there being some contingent properties. At least some of God’s properties, by contrast, are necessary. These are not subject to the principle of sufficient reason. So Kant clearly admits a distinction between the necessary and the contingent. This is important for the thesis, better known from Kant’s Only Possible Argument but already in place in 1755, that the material content of possibilia has its ‘source’ or ground in God (ND 1:396). Kant denies, however, that what is grounded in God in this way is ‘absolutely necessary’ (1:396). I take him to mean that what actually happens is contingent, even though God grounds the material content of all possibilities.

Again and again during this period, Kant refers to divine choice in creation, understood in terms of possibly doing otherwise. We’ve already seen Kant consider the possibility of God’s creating morally perfect but unfree agents. Elsewhere he writes that God could have chosen to perfect our intellects, or to restore decaying planetary orbits to a more perfect state (ND 1:408). Presumably such choices, though absolutely within God’s power, would have resulted in an overall worse world, and thus are morally impossible.

In the contemporaneous Universal Natural History, many passages are straightforwardly committed to alternative divine choices in creation (UNH 1:271, 1:311, 1:336, 1:345). The work also lays out a distinction between God’s free choices and consequences of these choices that are not free. For example, God’s creating various features of celestial bodies was ‘tied to the mechanical rules of motion’ and therefore not an ‘act [of] free choice’ (1:344).⁹ Even these unfree creative acts have their source in God. Here too, then, it appears that divine freedom of choice must be distinguished in terms of something beyond mere source spontaneity.

In sum, Kant takes seriously the possibility that divine creation could undermine our freedom. He rejects the Leibnizian response to this problem, without capitulating to necessitarianism. Instead, many of his discussions suggest that (a) God enjoys freedom of alternative possibilities in creation, thus that there is at least one agent with indeterministic spontaneity; and (b) God could have undermined our freedom or spontaneity by creating us differently, even though in such a scenario we would still be the sources of our actions.

2.2. Activity, determination, and the will

Let us turn to Kant’s account of free action. We’ve seen that it includes, as a necessary condition, some degree of source spontaneity. The motivations of free action are recognizably mine, originating for example in my understanding, and not merely in ‘external stimuli and impulses’ (ND 1:400).

But this does not settle how the will’s choices relate to motivation. Three main options are available. First, the will’s choices might be strictly determined by prior motivations—whether these are appetites or desires, or else based in the understanding and other theoretical faculties. Versions of this determinist position have been attributed, rightly or wrongly, to Aquinas, Hobbes, Gassendi, and Wolff.¹⁰ Intellectual or psychological determinism threatens to exclude indeterministic spontaneity by making volitions effects that, though not necessarily physical, are broadly natural. Furthermore, once non-volitional motivations are taken as sufficient conditions for action, one can ask about the sufficient conditions for these motivations themselves. If this series of sufficient conditions leads beyond the agent, source spontaneity also seems to be in jeopardy.

Second, voluntary choice might be entirely unconstrained by motivations, such that ‘actions are the product of chance’ (ND 1:402). This is the view Kant attributes to Crusius, perhaps based on the Crusian idea that it is in virtue of their existence that things are spatiotemporally determinate and located outside of thought. More precisely, as Kant reads Crusius, the sheer actuality or existence of a particular event ‘alone’ explains why this event, and no other possible event, occurs: ‘there is no need for a determining ground’ beyond the existence of token acts of ‘the free will’ (ND 1:397). The underlying idea seems to be that an event is necessary at the time that it takes place, insofar as in occurring, it excludes other possibilities. This was a common medieval view, already endorsed by Boethius in his Consolation and commentary on De Interpretatione. It was supposed to be consistent with regarding the future as open or contingent (Normore 2003; Pasnau 2020).

Kant’s characterization of Crusius’s position—leaving aside its accuracy—makes it hard to see how an act of free will could have any explanation stemming from the agent.¹¹ On Kant’s understanding, the appeal to the actuality of an event in the present is supposed to dodge questions about how past events or ‘grounds which are prior’ in time relate to the actual existence of a given action (ND 1:397). In this Crusian account, a present event or act is posited as self-explanatory. Its actual existence somehow explains what it is. But ‘such actions,’ Kant objects, ‘would scarcely deserve to figure among the prerogatives of intelligent beings’ (1:400). These actions are random or inexplicable, for they have no explanation in terms of volitional or motivational grounds. Kant’s Crusius is committed to actions that occur without any determination by prior ‘reason,’ ‘deliberation,’ or ‘motives’ (1:406).

Third, one could adopt a middle position, on which choice is partly but not fully grounded in prior motivations such as reasons and desires. Decisions would then not be fully predetermined. They would have substantive grounds, because they are partly grounded in a free act of willing. Not all would accept the coherence of this middle position, and I will not try to defend it here. But views of this kind are of major historical significance. As Thomas Pink (2004) has argued, this was the standard Latin medieval conception of free will, from Aquinas up until Hobbes. The basic idea is that voluntary action, as an exercise of the will, is neither the mere effect of motivational forces such as appetites and desires, nor of rational beliefs about what should be done.

Many medieval accounts, for instance from Scotus and Suárez, add that a voluntary action in accordance with a reason is metaphysically distinctive, and cannot be understood on the model of a mere effect in the material world. Matter and its natural powers were typically seen as more determined, less free, than volitions (Pink 2004: 131). To take Scotus’s example, the natural power of coldness qua coldness cannot warm, whereas the will can either perform a given act or its opposite (1997: 137–39). Note that on this sort of picture, examining the sources of an action may provide a reliable criterion for determining whether it is free. Free actions will not find their source merely in the powers of matter. But this is because the relevant sources are not fully determined, and not just because they are internal to the agent.

I will argue that many passages in the ND commit Kant to this sort of middle position. We can begin by considering his initial line of response to the Crusian position on freedom. This response is surprisingly concessive. Kant grants to Crusius that if an agent’s ‘act of free will’ actually ‘exists’ at a certain time, then this would exclude the agent’s doing the opposite at that time (ND 1:397).

Here we would expect a determinist to stress that this act of free will is also fully determined by temporally prior grounds. Instead, Kant agrees with the Crusian that the act of free will is ‘contingent’ (ND 1:397). The main point he stresses is rather that the ‘inner existence’ of the act of free will—I take this to mean something like its intrinsic properties—does not entail all the relations that this act stands in (1:397). In particular, the intrinsic properties of the act do not entail all of the act’s temporal properties, such as the time it began.¹² So these temporal properties are not determined by the mere existence of the act of free will, but by some external grounds. He takes this to be enough to refute Crusius’s position. The key conclusion here, though, is only that the act of free will has at least one prior determining ground (because its temporal, relational properties have at least one prior determining ground). This does not settle whether the ‘inner’ or essential properties of the act are so determined.

More broadly, Kant repeatedly considers a strictly deterministic picture of the universe in order to address the Crusian objection that the principle of sufficient reason is unacceptable because it entails determinism. Kant grants that determinism poses a serious threat to freedom (1:399, 1:403).¹³ One of his main lines of response to the Crusian objection is to repeatedly deny that free actions, whether exercised by God or by finite agents, result from ‘ineluctable necessity’ (ND 1:400). Given the context, which is a discussion of determinism, Kant is plausibly read as saying that free actions are incompatible not only with necessitarianism, but also with determinism.

Next, we can turn to more positive accounts of free willing in the ND. Kant is careful to distinguish ‘desire[s]’ and ‘reasons’—i.e., the conative and the cognitive—from our voluntary ‘striv[ing] … in conformity with’ desires or reasons (ND 1:403). The way Kant deploys this distinction appears to commit him to incompatibilist conclusions. For example, he explicitly distinguishes ‘necessitated’ actions, such as reflexes, from free actions that are merely ‘called forth’ by motives (ND 1:400). To be sure, a distinction between the sources of free action is in play here: a necessary condition for my action’s being free is its (partial) source in my rational will. But I take Kant to also be articulating a difference in the (in)determinacy of action. The distinction between being necessitated and being merely called forth concerns not the source, but the metaphysical status of action.¹⁴

Furthermore, in a somewhat neglected passage, Kant characterizes free action as involving the ‘spontaneous inclination of the will,’ stressing that ‘the power to perform an action is suspended in a state of indifference relative to each of the two directions in which it could realize itself’ (ND 1:401). Kant singles out the will, rather than mere ‘motives of the understanding,’ as the locus of spontaneity in free action (ND 1:401). Since acts of will are just as internal to an agent as are motives of understanding, the distinction he is drawing does not seem to concern source spontaneity.

Here Kant uses indeterministic, counterfactual language: at least in one stage of choosing, volition is indifferent between alternatives and ‘could realize itself’ in either of two ways (ND 1:401). In stressing that volition might either perform an act or its opposite (unlike powers that are bound to realize a particular alternative), Kant seems to be calling attention to the undetermined nature of volition, above and beyond the question of the source of action. Kant frames this as a ‘generally admitted’ view or datum that needs to be accounted for in an adequate account of the freedom of rational agents (ND 1:401).

Kant does immediately add some qualifications. Free action does not consist ‘in being carried away in all directions towards objects by some vacillating impulse’ (ND 1:401). But this need not imply anything more than a rejection of the radical liberty of indifference that Kant associates with Crusius. Kant wants to claim that there must be some ‘inner’ motivational ‘grounds’ for action, but this does not entail that the will is strictly determined (ND 1:401). In claiming that free action is not indifferent vacillation, then, Kant is not taking back his commitment to the ‘generally admitted’ view that free willing involves genuine alternative possibilities. So it does not seem that Kant propounds this view merely in order to refute it.¹⁵

Kant’s non-Crusian position is elaborated in a dialogue between Caius, who takes actions to be in some sense random and groundless, and Titius, who apparently defends Kant’s position. Titius argues at length that the liberty of indifference is in some sense an ‘illusion’ (ND 1:403). At first glance, this looks like a rejection of incompatibilism. Yet in the history of discussions of free will, the term indifference is often ambiguous (Sleigh, Chappell, and Della Rocca 1998: 1210). While medieval conceptions of indifference typically stressed the metaphysical availability of alternative possibilities, later thinkers such as Descartes took indifferent actions to be those an agent performs for no reason. Such ambiguities also appear in Leibniz. He denies that rational agents ever act without a reason (1875–90: III:168; 1989: 195). Yet immediately after rejecting this sort of ‘equilibre’ or equipoise, Leibniz upholds what he calls ‘liberty of indifference’; so by the latter he cannot mean action without any reason (1875–90: VI:128).

Now when Caius claims that agents are in a radical ‘state of indifference’ between acting one way and another, he grants that this notion of freedom would ‘eliminate’ ‘all … grounds’ of free action whatsoever (ND 1:402). So in taking issue with Caius, Titius need not reject the metaphysics of libertarian freedom as such. Titius can be read as endorsing a conception of rational volition on which voluntary actions are not indifferent, in the sense of being merely random or a matter of chance. The action is grounded in volition and in (non-determining) prior motives. As Kant has Titius say:

Since grounds attract us in a certain direction, we shall, in order at least to test our freedom, turn our attention in the opposite direction, and thus make it preponderant so that the desire is directed thus and not otherwise. In this way, we shall easily persuade ourselves that determining grounds must certainly be present. (ND 1:403, emphasis in original)

The final phrase in this passage stresses the way in which action is determined by grounds. But this is downstream, as it were, from the volitional locus of freedom: we ‘make’ up at least some of these grounds voluntarily.

Elsewhere, Titius objects that Caius’s conception of freedom threatens to remove any role for prior motives, and especially for virtues of character, which Titius seems to understand as pre-existing settled dispositions (1:402). If actions are really performed with freedom of indifference, then they will not be based on virtuous dispositions but on inexplicable ‘baleful fate’ or ‘chance’ (1:402). Titius’s speech is at a high rhetorical pitch. It ridicules Caius’s position as counterintuitive and self-undermining. This style contrasts with Kant’s earlier neutral or positive assessment of the ability to do otherwise as a ‘generally admitted’ datum. So both the content of Titius’s objections and their tone indicate that Kant draws a distinction between the metaphysics of libertarian freedom as such, and a Crusian version of libertarianism that entails indifference.

To illustrate the account of agency that Kant has Titius advocate, consider a case where I am choosing among two desirable but mutually exclusive options. For example, I am planning to have dinner, but decide to go for a walk instead. Titius does not describe this as a case where my antecedent motives strictly determine that I will go for a walk. Instead, the idea is that I will to make the desire to go for a walk ‘preponderant’ over the desire to dine that had previously ‘attract[ed]’ me, and so act in accordance with the former desire (ND 1:403).

My going for a walk is not random or indifferent. Rather, it is ‘determined by grounds.’ It is grounded at least in (1) my desire to go for a walk and (2) my willing to act in accordance with that desire, rather than the desire to have dinner (ND 1:402). Here (2) is an act of agent causation not fully determined by prior motives.

Should this distinction seem overly subtle, it is worth noting that Baumgarten—despite his reticence about the metaphysical issues involved in freedom—also clearly distinguishes two faculties associated with (1) and (2). Baumgarten takes arbitrium as a faculty of ‘desiring and averting’ (that is, a faculty that determines motivations) to be distinct from libertas as a faculty of ‘willing or refusing’ (2013: §712, §719). Kant flags this distinction in notes on Baumgarten from around 1755, classifying sensory appetites as examples of mere arbitrium (Belieben) and indicating that non-human animals as well as humans possess this faculty (Kant 2019: 277, 280). Both of these faculties are in turn distinct, on Baumgarten’s account, from mere spontaneitas in the sense of being the source of one’s actions.

I take the moral psychology of the ND to also invoke (3) a capacity to attend to a given desire rather than some other, conflicting desire. Kant stresses, for example, that ‘turn[ing] our attention’ is an exercise of ‘freedom’ (ND 1:403). This capacity originates in the will: genuine choice involves willing our intellectual faculties to reflectively consider our options before acting. Options can then be made more or less salient.¹⁶ Cases of akrasia and other kinds of practical irrationality can be partly explained in terms of this capacity. Through selective attention, the lesser good may seem to be to be a better choice than the greater; one may even come to entirely ignore a good.

These discussions of agency and moral psychology are importantly linked to Crusian anxieties about determinism. Caius worries that if determinism is true, his responsibility for his own ‘misdeeds’ does not fall on him, since his actions are fated from creation (ND 1:401). The worry, as he puts it, is that ‘every inclination of my will has been completely determined by an antecedent ground … right back to the beginning of all things’ (ND 1:402).

One aspect of Titius’s response to this worry is to stress the importance, in the context of moral responsibility, of an ‘inner principle’ as the source of one’s actions (ND 1:402). This has been taken to commit Kant to a compatibilist position. But that does not follow. On the one hand, it is possible to acknowledge that source spontaneity is crucially involved in moral responsibility, for example in practices of praise and blame, without taking it to be sufficient for free action. On the other hand, we’ve seen that Kant’s interactive metaphysics does not permit actions to originate from an absolutely inner principle.

The first response Titius makes to Caius is in fact straightforwardly concerned with determinism as a threat to the alternative possibilities required for freedom. Titius, speaking for Kant, says:

At any given juncture, the series of interconnected grounds furnishes motives for the performance of an action which are equally attractive in both directions: you readily adopted one of them because acting thus rather than otherwise was more pleasurable to you. But you say: it was already determined by the totality of grounds that I should incline in one particular direction. I should, however, like you to consider whether it is not the case that the spontaneous inclination of your will, according to the attractions of the object, is not required if there is to be a complete ground of action. (ND 1:402)

At the moment of choice, on this picture, the agent has available two ‘equally attractive’ sets of motives. But neither set of motives is a sufficient ground of action. Instead, a further ground must be added, which originates in the will and is not determined by either set of motives. This further ground is a necessary condition for acting in one way, and not the other.

So this passage apparently commits Kant to an incompatibilist account of freedom, on which the way we act is not determined by pre-existing motives, but is also necessarily grounded in the inclination of the will one way or the other. No matter how much ‘the state of things prior to … free acts has been determined,’ the ‘hinge’ of action remains ‘voluntary’ (ND 1:404). Again, going for a walk, as an action based on volitions, may be determined: the will plus all other grounds are jointly sufficient conditions for it. But the volitions themselves need not be determined, even if they are influenced by, or partly grounded in, desire for ‘the attractions of the object.’ While possible objects of choice provide the matter for willing, a further necessary condition for settling choice originates in the agent’s will.

Let us turn, finally, to some passages where Kant can seem, contrary to my reading, to endorse determinism and compatibilism. These passages have been taken to rule out the reading of the ND I defend here. But I’ll argue that they are far from conclusive. In one such passage, Kant writes:

In so far as they [sc., the free actions of human beings] are regarded as determinate, their opposites are indeed excluded; they are not, however, excluded by grounds which are posited as existing outside the desires and spontaneous inclinations of the subject, as if the agent were compelled to perform his actions against his will. … On the contrary, it is in the very inclination of his volitions and desires, insofar as that inclination readily yields to the blandishments of his representations, that his actions are determined by a fixed law and in a connection which is most certain but also free. (ND 1:400)

To begin, note that Kant’s topic is the determination of action by the will, not the determination of the will itself. And his framing of the issue is qualified. Free actions are considered ‘insofar as they … are regarded as determinate.’ Taken literally, this implies that actions could also be regarded as indeterminate.

Even the references to a ‘certain’ connection or a ‘fixed law’ at the end of the passage need not entail strict determinism regarding the will. For this claim is in the first instance about action, which is causally downstream of acts of willing. Additionally, we have seen that in Leibniz and his medieval predecessors, a high degree of certainty is compatible with contingency. Finally, it cannot be assumed that a ‘fixed law’ means a deterministic natural law. For example, moral necessitation might be lawlike without being strictly deterministic. I return to this issue in section 4 below.

While the passage from 1:400 aims in part to defend source spontaneity, I take it to suggest that indeterministic spontaneity is required as well. This lends support to my reading. Consider the first sentence of the passage. On the conception of source spontaneity laid out here, free actions are not grounded in what exists ‘outside the desires and spontaneous inclinations of the subject.’ Kant, however, has shown himself to have grasped Crusius’s objection that if strict determinism were true, the ultimate grounds of free actions would lie outside the subject’s will and spontaneous inclinations. A few pages later, he reiterates that moral responsibility requires that agents’ actions ultimately originate in an ‘inner principle of self-determination’: in the agents ‘themselves,’ rather than a causal chain that merely passes through them (ND 1:404). Kant’s interactive metaphysics, if deterministic, would undermine this kind of ultimate origination of free actions in agents. So there is a way to read this passage as holding that if one takes source spontaneity seriously, as ultimate responsibility, one will be led to see that it requires indeterministic spontaneity at some point in an agent’s history.

Another potentially problematic passage appears in the dialogue between Titius and Caius. In one exchange, Titius may seem to accept determinism and hence compatibilism, implying that Kant does so as well. Caius objects that on Titius’s account of moral psychology, the ‘will could not have failed to incline’ towards one action rather than another, and Titius responds that this does not eliminate spontaneity or freedom, if freedom is properly understood (ND 1:402). However, especially read in the context of Titius’s broader goals in the dialogue, this passage need only commit Kant to the idea that some prior motives (which ‘incline’ it to some extent) were predetermined, rather than that volition is strictly predetermined. While Titius does emphasize here that spontaneous action ‘issues from an inner principle,’ this can accord with a reading on which source spontaneity is a necessary but not sufficient condition for freedom (1:402). Here too, Titius can be read as rejecting a specifically Crusian view rather than possible incompatibilist accounts.¹⁷

A final important textual reason why the early Kant has been read as compatibilist is his defense of a version of the principle of sufficient reason. The principle, which he prefers to call the Principle of Determining Ground (PDG), might seem to guarantee that every action is causally determined. And there is a long history of conflict between libertarian freedom and explanatory principles such as the principle of sufficient reason. Murray (2004: 8) contends, for example, that this sort of tension can already be found in Aquinas. I will argue in the following section, however, that Kant’s specific version of the principle of sufficient reason does not exclude an incompatibilist reading.