On the Transcendental Freedom of the Intellect

3.1. Practical & Transcendental Freedom

We can understand practical freedom as follows:

Practical freedom is typically demonstrated by rational agents pursuing ends even when they conflict with the agent’s immediate inclinations or interests. Rational agents are capable of pursuing ends in the face of contravening desires or competing pleasures. As Kant points out,

The famous “marshmallow tests” examining the extent of a child’s capacity for delaying gratification, as conducted by Walter Mischel in the 60s and 70s, can be understood as testing the extent or strength of this capacity for practical freedom.[14]

While Kant is clearly talking, in the Canon quote above, about the capacity, independent of current competing desires and inclinations, to form intentions for carrying out bodily actions, much the same point can be made concerning intellectual acts of the kind discussed in the previous section. Rational adults exhibit, at least some of the time, an ability to conceptualize, judge, infer, and ultimately assent on the basis of something other than what the subject currently wants, desires, or otherwise has an immediate inclination to act on, and often contrary to what the immediate deliverances of sense would seem to indicate.[15] In particular, such subjects can engage in intellectual acts that respect canons of evidence and laws of truth-preservation even if such canons and laws yield verdicts contravening one’s current interests, inclinations, or the strength and character of one’s sensations. Hence, the core notion of “practical freedom” is just as significant for theoretical uses of the intellect as it is for the more traditionally discussed practical uses exemplified by acts such as willing or intending.

Are there further forms of freedom that a subject must exhibit in order to possess practical freedom? In particular, does practical freedom require what Kant calls “transcendental freedom”? Kant discusses the notion of transcendental freedom in the resolution of the Third Antinomy. He says,

So, for Kant, the idea of freedom, understood as “transcendental freedom”, is the idea of an “absolutely spontaneous” act (A448/B476)—that is, an act whose ultimate causal source lies in the subject, and is not itself determined by any temporally structured causal ground.[16] Kant denies that we can ever know whether an act is free in this sense (A557/B585). Kant also denies here that practical freedom is wholly independent of transcendental freedom; the practical conception is “grounded” on the transcendental, and he subsequently says that “the abolition of transcendental freedom would also simultaneously eliminate all practical freedom” (A534/B562). This seems to be Kant’s considered view, for in the discussion of freedom in the second Critique Kant also claims that without transcendental freedom “no moral law is possible and no imputation in accordance with it” (5:97).

However, the conception of transcendental freedom so presented can at least seem to conflict with the discussion of freedom in the Canon of the first Critique, for Kant says in that late section that “Practical freedom can be proved through experience” (A802/B830), in that normal adult human beings clearly have practical freedom in the sense I defined above. Kant’s claim in the Canon presents a straightforward problem, for if practical freedom depends on absolute spontaneity/transcendental freedom, and we cannot know we have the latter, then we cannot know we have the former, and thus cannot “prove it” through experience.[17]

What to make of the text of the Canon then?[18] One way to try and resolve this plays off of Kant’s claim in the Canon that

One reason Kant might say that we can simply “leave aside” the issue of transcendental freedom when dealing with the practical, is that at time of writing the Canon Kant really did conceive of them as independent of one another. Hence, on this way of explaining the text, the conception of practical freedom in the Canon is an early statement of Kant’s view concerning the relation between transcendental and practical freedom, one which he ultimately revises in the statement made in the discussion of the Third Antinomy (which was written after the text of the Canon, despite coming before it in the work), and which is subsequently given its clearest expression only in the second Critique.[19]

Another possibility is that Kant utilizes both a stronger and weaker sense of “practical freedom” in the Canon.[20] The weak sense is merely that of being able to resist immediate sensible impulses. This is something, as we’ve noted, of which normal adult humans are clearly capable, and is thus provable via experience. What is not provable through experience is possession of practical reason in the stronger sense—that is, that a causally ultimate faculty of reason explains the possession of the capacity for practical reason in its weak sense in normal adults. Kant may be indicating that it is this stronger sense that can be put aside when he says,

If it is merely nature that ultimately explains the weak sense of practical freedom, then we lack practical freedom in the strong sense, for we lack the capacity for reason to be casually ultimate in determining our wills. Hence, only practical freedom in this strong sense—as requiring determination by reason rather than nature—would depend on transcendental freedom.

In any case, whether Kant developed his view after writing the Canon into the more resolute view of freedom on display in the Third Antinomy and the second Critique, or he consistently maintained his view from the writing of the Canon, the result is the same—according to the mature critical philosophy rational determination of one’s actions presupposes transcendental freedom. I understand Kant’s conception of transcendental freedom as follows:

A transcendentally free rational agent must be the causally ultimate source of her actions—call this the “source requirement” for transcendental freedom. That is, the agent must be able to (i) initiate a causal series, which includes or constitutes her act, without being (ii) causally determined in this initiation by any temporally structured causal ground.[21] Moreover, since Kant locates the source of transcendentally free action in the agent’s power to so initiate a causal series, it cannot be the case that transcendentally free acts just happen as wholly random or indeterministically caused events in nature. As Kant notes of the view of Lucretius, such events or happenings would amount to “blind chance” (CPrR 5:95).

Hence, for Kant, transcendentally free acts are not indeterministic, they are causally determined. It is just that they are causally determined by the agent and not by the mechanism of nature. This is particularly clear from Kant’s discussion of “predeterminism” in the Religion,

Absolute spontaneity is not at risk from determinism but from predeterminism. Satisfaction of the source requirement thus depends on the agent’s actions not being exhaustively causally determined by preceding events.[22] Given the causal principle articulated in the Second Analogy, the natural world is such that, for any agent, the ultimate determinants of her actions would be in the past, and thus beyond her “control” (Gewalt).[23]

As Kant conceives of the causal order of the phenomenal world, it is such that within this order, one never has control over one’s actions, and this lack of control is sufficient to undermine one’s freedom. Transcendentally free action within the phenomenal realm is thereby impossible (I discuss exactly why in the next subsection). Transcendental freedom thus requires a different causal order, one whose possiblity depends on features of negative noumena or things as they are in themselves.[24]

It is worth emphasizing here that, as I understand Kant’s conception of free action, it depends on the coherence of his distinction between things as they appear and things as they are in themselves, and thus on his transcendental idealism. The exact meaning of Kant’s doctrine of transcendental idealism is, however, contested. I take it though that in determining the correct interpretation of Kant’s idealism we are constrained by the conditions Kant sets on the kind of control outlined here—viz. the metaphysical conditions on the ultimate ground of explanation of a rational action. The correct interpretation must allow for a logically coherent distinction between a temporal and a non-temporal causality, with it being at least epistemically possible that rational agents possess such a non-temporal causal power. However, I leave open here whether, or to what extent, this conception of freedom is compatible with the wide array of readings of Kant’s idealism, from purely epistemic “inference ticket” readings like that of Henry Allison, to more robustly metaphysical readings like those of Ameriks, Langton, and Allais.[25]

3.2. Transcendental Freedom & Control

We’ve seen that Kant construes transcendental freedom in terms of a rational agent’s causal ultimacy—what he calls the agent’s “absolute spontaneity”—and that he thinks of the temporal causal structure of the phenomenal world as threatening such ultimacy by placing the causal grounds of action beyond the agent’s control. But it is not entirely obvious why the temporal structure of phenomenal causation threatens the control necessary for transcendental freedom, or exactly how we should understand the notion of control at issue.

There seem to be at least three reasons why Kant rejects the compatibility of temporal causality with control—viz. that (i) it entails the existence of a causal ground distinct from and independent of the agent; (ii) it entails the existence of a ground over which the agent can exert no causal influence; (iii) it entails the operation of a form of causality that is not characteristic of the activity of the agent’s intellectual faculty. Kant does not always distinguish between these three reasons, but I think that while he often appeals to either (i) or (ii), in fact it is (iii) that is the root of the incompatibility between temporal causation and control, and which also explains his endorsement of (i) and (ii). Let me explain these in turn.

First, if the ultimate causal grounds of action can be traced to temporally remote causes that lie outside the agent, then it is clear why she would not be transcendentally free—her act would not be initiated by her but by something else. Note here that it is not time per se that is the ground of her lack of freedom, but rather the fact that the ultimate cause was external to her. Time would be an aid to determinism but not itself directly undermining of the subject’s control. Something like this issue seems to be at work in Kant’s discussion of theological determinism and the transcendental reality of time in the Critique of Practical Reason. There he raises the worry that,

There are a number of points made in the passage. Kant first suggests that if we are creatures of God then there seems to be a problem for freedom. Moreover, if time is transcendentally real and God is the ultimate causal origin of the existence of all things in their complete determinacy then Kant further concludes that it would be unavoidable that all beings have their acts ultimately causally grounded in God. In acting on the basis of causal grounds determined by another, humans would thus be “marionettes,” controlled by an external source.[26] This is even more explicitly articulated in the lectures on ethics:

Hence, on this conception of the problem of (pre)determinism, time is merely its handmaiden. It is God’s causally creative activity that provides the real threat to control, and the reality of time would make this threat unavoidable by making dependence on an external causal source for one’s acts unavoidable.

However, there are (perhaps broadly Leibnizian or Spinozistic) conceptions of the causal structure of change in a subject’s states that are temporal but do not (or need not) appeal to causally external factors. In such cases the causal ultimacy characteristic of transcendental freedom would seem to be present, for no cause is appealed to from without the agent, even though each of the agent’s states is temporally ordered, and the ground of the change between states exists at a preceding time. Does Kant then lack any basis for rejecting a position that allows for temporally preceding causal grounds, so long as they are all within the agent?

In fact, Kant objects to just such a position as failing to put the levers of agency under one’s control. Kant contends that such a position articulates a merely “relative” spontaneity—a “comparative” sense of freedom that he deems a “wretched subterfuge” (CPrR 5:96). Kant dismisses the possibility that this comparative sense could substitute for the control of one’s actions required by morality.

Here Kant’s objection to the merely comparative freedom of being determined by one’s internal states rather than an external force is that these internal states would still be in time, and thus that the determining grounds of one’s act would lie in a previous time, and thereby outside of one’s control. Note that here the argument does not rely on the way in which temporal causality traces back to an agent-external causal ground. Rather, the problem is temporal causality itself. Kant goes on to make this point quite clearly:

Kant claims here that it is specifically the temporality of phenomenal causation that threatens the agent’s control, and thus that even when the determining grounds of an agent’s acts are all internal to the agent, the fact that they are in time is inconsistent with the agent’s control with respect to them—the agent has no more freedom than a “turnspit” (Bratenwender, 5:97; see also Metaphysik L1, 28:268–69).

But why is the temporal structure of causation itself a threat to the agent’s control over her actions?[27] After all, on the above model the agent is, in every action, the ultimate causal source of that action, even though this causality is, as it were, “spread out” over the temporal history of her activity. Moreover, on this model there is no clear external threat to her agency in the manner that God’s creative causal act of determining her nature and existence is an external threat.

One possible answer is that Kant is pointing in these texts to a second issue, distinct from that of agent-external determining grounds. Because of the nature of time, an agent acting at time t_n cannot causally affect the grounds g of that action, which exist at some prior time t_n-1, because (finite) agents cannot causally affect the past.[28] Hence, in the above passage Kant may be arguing that, putting aside the issue of God’s creative activity, so long as time is transcendentally real one would lack freedom because one’s acts would be due to grounds that are not within one’s causal reach or influence at the time in which one acts.

However, one might worry about this being a compelling objection to a broadly Leibnizian (or Spinozistic) position, since at no time is the subject caused to act by a ground that does not lie within the subject. While it is clear that finite agents have no causal influence on the past, Kant does not give us further explanation of why this would undermine the central issue of the agent’s control if all the relevant grounds of action are within the agent. Hence, it is not clear that emphasis on a lack of causal influence on the past provides a non-question-begging means of countering the position that freedom is compatible with determination by temporally preceding grounds.

I think we should thus consider an alternative reason for rejecting the compatibility of temporal causation with control, though this will require a brief excursus on the issue of Kant’s conception of control over one’s acts. Recall from Section 2 that Kant thinks of an action (Handlung) in its most general sense as belonging to the power or capacity of a substance to cause a particular accident (Metaphysik Pölitz 28:564–65). He then conceives of each faculty, whether receptive or spontaneous, as acting in its own specific way by producing specific characteristic effects, and thus operating according to specific and unique casual laws. The spontaneous acts of the intellect follow laws of “combination.” Such acts are a form of “self-activity” (Selbsttätigkeit) and can never be given through the intuition of an object.

We saw in Section 2 that Kant’s conception of a substantial act is very broad, encompassing any actualization of a substance’s causal powers through an exertion of force to bring about an effect (a property or “accident”) in itself or another being. While a substance’s acts thus explain all of its accidents, Kant nevertheless wants to maintain the intuitive distinction between what happens to a substance versus what the substance does. He needs such a distinction, moreover, to distinguish between the two stems of our cognitive power—“passive” receptivity and “active” spontaneity.[29] He accomplishes this distinction by analyzing the relevant cognitive faculty or power in terms of whether or not its exercise requires appeal to a “determining ground” whose nature or existence is distinct from that of the power in question, or whether the determining ground is itself part of the power’s exercise.[30] ,[31] Concerning acts that a subject does of itself, there may be further questions regarding whether those actions are voluntary or involuntary, which would then require understanding whether and how the will or power of choice is determined. But given this distinction between what happens to a subject and what the subject does we can speak (clearly if perhaps somewhat awkwardly) of “passive” or “receptive” acts of a subject in contrast to its “active” or “spontaneous” acts. For example, the sensory capacities individuated in characterizing our form of receptivity (e.g., vision, touch, etc.) all must appeal to determining grounds distinct from any exercise of those capacities.

In contrast, in the “active” acts of our spontaneous intellectual capacities, the existence and nature of the determining ground of the act is not itself independent of any exercise of the intellectual capacity. For example, in an inference, the determining ground of the inference, insofar as it is a rational act, must be the recognition of a logical relation between the premises, which is to say that the nature of the determining ground of the inference—that is, its status as a determining ground—itself makes reference to the very capacity for inference whose actualization is in question.[32]

On this way of understanding Kant it should be clear why he would typically construe spontaneous mental activity as an action—in the sense of something an agent does—since the intellect operates in such activity according to its own laws, and the determining grounds for its actions are themselves inextricably related to the nature of any exercise of the intellectual capacities themselves.

With this characterization of the nature of the receptive and spontaneous powers of the rational subject in hand, we can get a clearer sense of Kant’s notion of “control” and its incompatibility with temporally structured causation.

As I understand Kant’s notion of “control” (Gewalt), a subject is in control of their acts when the determining grounds of those acts are not independent of any exercise of the capacities for so acting. This means that an agent cannot be in control of any act that is itself the outcome of a determining ground whose existence and nature is itself independent of any exercise of the relevant capacity. Hence, all receptive acts are outside the agent’s control, which constitutes the sense in which they are passive “happenings” to the agent and the way in which something is “given” to it.

We can now see why Kant would reject the compatibility of the controlled exercise of a rational, spontaneous capacity with any temporal causal ground of its exercise or actualization. It is essential to the nature of a temporally structured determining ground that its position in time play a decisive role in the function of the ground to determine its particular effect. Hence the schematized law relating ground to consequence (as cause to effect) is one that makes essential reference to a temporal relation, and thus the position of each relatum in time. But then, even if the determining ground lies entirely within the agent, the sheer fact that it occurs in time means that the nature of that ground is, at least in part, independent of any exercise of the subject’s intellectual capacity or faculty, which itself makes no such reference to temporality. Thus, to construe a subject’s acts as brought about via a temporal causal ground is to construe those acts as, by definition, brought about via “alien influences” (fremden Einflüssen; G 4:448), that is, through grounds and laws external to one’s spontaneous capacities.[33] Moreover, since we distinguished the active and passive powers (i.e., receptivity and spontaneity) above in terms of whether the causal ground for their exercise ultimately depends on the relevant faculties, if an act is the result of a determining ground in time, then that act is due to an exercise of the subject’s passive receptive power.

Kant’s position thus stands as a genuine counter to the Leibnizian one rather than simply begging the question against it. For Kant articulates a seemingly inescapable outcome of Leibniz’s conception of the temporal structure of a substance’s acts—viz. that since the determining grounds of the substance’s acts are always temporal, the acts thereby necessitated will be receptive rather than spontaneous.[34] Put another way, this counter explains why Kant thinks the Leibnizian cannot recognize an absolute moment of spontaneity, but at best only the relative spontaneity of temporally preceding determining grounds that lie wholly within the subject but which are independent in nature and thus alienated from any exercise of its rational capacities.[35]

Moreover, given this conception of control we can see why Kant would regard causation by an external source or entity (e.g., God) over which one has no causal influence or reach as threatening to control. Both the case of agent-external determination and the case of agent-internal temporal determination are instances of determining grounds that exemplify an independence incompatible with the actualization of an agent’s spontaneous rather than its receptive capacities.

We can also now more fully appreciate why Kant defines transcendental freedom in the manner that he does. Recall that transcendental freedom consists in the (i) power to begin a causal series from oneself; (ii) without being determined to do so by any temporally structured causal ground. How should we understand the relationship between Kant’s conception of control and his conception of transcendental freedom? One has control over one’s acts just in case one is transcendentally free, but the explanation of why Kant conceives transcendental freedom in the manner he does stems from the conception of control I have articulated. When one has control one exercises a capacity determined by a ground that depends on the very capacity so exercised. Hence one must be the causally ultimate source of the act. And for the reasons elucidated above the control one has in exercising such a capacity is incompatible with the ultimate determining ground of that exercise being in time. Moreover, the centrality of the notion of control for Kant’s overall conception of freedom also explains why he moves in the various texts I quoted above from conditions that undermine control to a lack of (transcendental) freedom.

3.3. Leeway & “Ought”

Thus far we’ve seen Kant’s conception of freedom to be essentially causal, and concerned with the ultimate source of the causality of action, as well as the kind of control that the agent thereby exerts over those actions. However, one might presume from Kant’s discussion that transcendental freedom also essentially involves the ability to do otherwise, that is, it essentially involves, in addition to the source condition outlined above, a leeway condition. Certainly, such a leeway condition is closely linked to Kant’s conception of the power of choice (Willkür). One feature of Kant’s definition of this power of free choice is that it is “a power to do or to refrain from doing as one pleases” (MM 6:213). Nevertheless, the concept of transcendental freedom does not entail leeway. A passage in the Religion, from which I’ve already partially quoted above, indicates this.

As this passage shows, Kant does not conceive of the sort of transcendental freedom constituted by absolutely spontaneous action as requiring leeway in the performance of those actions. God’s acts, including his acts of will, cannot be other than as they are, but this does not mean, in Kant’s view, that God’s actions are unfree. Despite lacking leeway in how he acts, God is absolutely spontaneous in the production of his actions, they are always within his control, and he is thus transcendentally free. Hence, transcendental freedom in general does not essentially involve leeway or the ability to do otherwise. Kant is therefore a source rather than a leeway incompatibilist: the key notion of (transcendental) freedom is not the ability to do otherwise, but to be the undetermined causal source of one’s actions, insofar as those actions are under one’s control.[36]

For Kant there is nevertheless an important connection between transcendental freedom and leeway, for it is precisely the fact that we can control our actions that makes possible the ability to do otherwise. Moreover, leeway of action is a necessary condition for the applicability of “ought” principles to us, and for us rational laws are presented via a system of “ought” principles. Kant says of the principles determining the will that,

Since the will (and ultimately the power of choice or Willkür) of a finite rational subject may or may not be determined by reason, the finite subject experiences the intellectual laws that might determine it as imperatives, and the application of such a law as a kind of necessitation. Since God is always perfectly determined by his rational intellect, and thus lacks any leeway in his actions, imperatives are correspondingly lacking.

One thing that is important to note in Kant’s conception of the conditions under which imperatives apply to rational subjects is that the relationship between the applicability of an “ought” claim and leeway need not require that choice always be directly operative in the occurrence (or not) of a specific act. For example, there are norms governing how one may adopt a particular doxastic attitude—what Kant calls holding-for-true or assent (Fürwahrhalten)—even though Kant denies that we can directly choose, irrespective of the epistemic grounds, whether or not to opine, believe, or know the truth of a judgment (JL 9:74; see also Logik Blomberg 24:158). Since leeway is a condition of the applicability of a norm, we possess leeway with respect to such attitudes not in the sense that we directly choose them, but rather in the sense that, if we are rational, then it is possible that such attitudes can be causally determined by rational rather than sensible causes.[37] One can thus “make up one’s mind” based on rational grounds rather than merely how one is sensibly determined. If it weren’t possible for the subject to adopt an attitude of assent based on rational causes then no ought would be applicable, and the rational norms would have “no significance whatever” (A547/B575).

Kant also endorses an “ought implies can” principle; e.g., “from the practical point of view this idea [of a prototype of humanity pleasing to God] has complete reality within itself, for it resides in our morally-legislative reason. We ought to conform to it, and therefore we must be able to” (Rel 6:62). If normative principles apply to us, it must be that we are, at least in principle, able to act in accord with them.[38] Hence, if one can never do otherwise because one is entirely (naturally) causally determined in one’s acts, it would be false that one ever ought to do otherwise. In this way the thoroughgoing determinism of the natural world threatens the applicability of “ought” judgments, understood as imperatives. If the thoroughgoing mechanism of nature entails that it is never true that one can do otherwise, there would be no point of a system of “oughts”, moral or otherwise. Hence, in order that “ought” statements for a subject be true at all, they must be true not only when the subject does x, but also when she fails to do x.[39] So, by Kant’s argument, “ought” principles cannot hold or be true if a subject could not have acted otherwise than she had. In order to have acted otherwise, however, it is not enough that the course of events leading up to and including her action could have turned out differently, in the sense that chance events might have occurred differently. Instead, an agent must have a capacity for action that is not subject to causal determination by factors beyond her control.

In sum, Kant construes free action as possible only if an agent has the requisite kind of control over her acts characteristic of the operation of her spontaneous intellectual capacities. Further, in finite rational agents, the applicability of norms as imperatives (i.e., as a system of “ought” claims) requires not only that the agent’s acts constitute or are otherwise a part of a causal sequence originating in the controlled exercise of spontaneous capacities, but also that she possesses leeway in her acts, with this latter fact explained by the agent’s being the ultimate causal source of her action.[40] Given these two conditions—viz. that an agent be transcendentally free in that she is the appropriate source of her action and that she have the leeway in so acting necessary for being subject to normative principles formulated as “ought” claims—Kant’s conception of the temporal structure of phenomenal causality threatens the possibility of the truth of one or both of these claims.

The view thus far articulated explicates the incompatibility between Kant’s conception of a controlled use of one’s faculties and his conception of the mechanical temporal causality of phenomenal nature. However, a number of questions remain. For example, to what extent are the activities of the intellect really free, especially in their theoretical use and what is the relation between such activities and will or choice? What reason is there for thinking that human beings instantiate the kind of control required for transcendental freedom? Even more generally, why think that such a conception of control is even possible for beings like us? In the next section I look at several different arguments concerning the transcendental freedom of the intellect’s activity that employ the notion of cognitive control outlined here, and which attempt to provide answers, or at least the outline thereof, to such questions.

4.1. The Imputability Argument

By far the most discussed argument as to why Kant is committed to some form or other of transcendental freedom in regard to even theoretical uses of the intellect is what I will call the “imputability argument.”[41] According to this argument, imputation of acts to a being such that they may be held responsible for those acts, and may thus be appropriate objects of praise, blame, and other “reactive” attitudes, requires that such beings be transcendentally free.

We see this in Kant’s distinction between ascription (zuschreiben) and imputation (zurechnen) in the following lecture transcription.

In the lectures Kant uses the example of a drunkard whose actions while inebriated are not strictly imputable to him but merely ascribable (Moralphilosophie Collins 27:288). What is imputable to the drunkard is the decision to become (highly) intoxicated in the first place, for it is this action that Kant construes as under the drunkard’s control, and correspondingly it is this action for which the drunkard can be held responsible, and to which his other actions while drunk must all be related. In the later Vigilantius ethics lectures Kant makes this point explicitly.

To the extent that the drunk was the “author” of his act—that is, made a free choice in deciding to get drunk—then the acts that are, strictly speaking, merely ascribable to him while thoroughly intoxicated are going to be things for which he is responsible because of the origin of those actions in the imputable decision to get drunk. We can thus say that the drunken actions are indirectly imputable to the drunk in virtue of this original act, but his blameworthiness for these acts is going to be directly dependent on his responsibility and blameworthiness for this original act.

At root here in the distinction between ascription and imputation is the notion of control. For Kant, ascription requires merely that one be the locus of some behavior or event, the determining grounds for which may be the laws of nature or something else entirely, as with a coerced or manipulated person. In such cases, we do not treat the actions of the coerced or manipulated (e.g., “brainwashed”) as under their control, and thus as appropriate targets of reactive attitudes or rational criticism. Similarly, non-rational animals may have acts ascribed to them but not imputed, for they are caught up in the deterministic mechanism of nature, which places the ultimate causal grounds of an act outside the agent’s control.

As we saw in the previous section, Kant considers control a necessary condition of free action so it’s unsurprising that he indicates its relation to the conditions of imputation. The epigraph to this paper records Kant as saying that

Since we’ve seen that Kant’s mature critical view considers control to be incompatible with one’s acts being the result of temporally conditioned grounds or causes, it is clear that Kant assumes such control to be possible, if at all, only at the noumenal (i.e., non-temporal) level. One can therefore summarize the imputability argument as follows:

1. (Non-proleptic) praise/criticism of an agent for their actions is only rational if the agent is responsible for their actions.[42]
2. Responsibility for one’s acts requires that those acts be imputable to the agent rather than merely ascribable.
3. Imputation of action requires that an agent have the requisite control over her actions.
4. Control of action necessary for imputability entails the transcendental freedom of that action.
5. ∴ Non-proleptic imputation and related praise/criticism of an action requires that the agent performing the action be transcendentally free.

There are at least two related worries concerning the above argument. The first is that considerations of imputability apply only to voluntary actions, and it is not at all obvious that intellectual acts, especially acts of synthesis, are voluntary. For example, it is plausible that Kant does not construe all acts of synthesis or concept formation as under the direct control of the will (Wille) or the power of choice (Willkür).[43] This worry thus affects the plausibility of premise (3), for one might object that the notion of “control” at work here requires the possession and relevant exercise of a will. If that’s correct then only voluntary acts can satisfy Kant’s conditions for imputability. The second worry concerns whether the imputability argument is suitable for showing a connection between imputability and specifically theoretical (as opposed to practical) exercises of intellectual freedom. This worry concerns the scope of the conclusion—can it apply to theoretical acts as well as practical ones? I’ll take these in turn.

First, does imputation require that the relevant act always be under voluntary control, and thus a product of the exercise of the agent’s will, or can even non-voluntary acts be under one’s control?[44] More specifically, the question is whether imputation requires volition or an act of will at all.[45]

In my view there is no specific text of Kant’s that definitively settles this issue one way or another. Instead, we need to work out an account that best fits with his explicitly expressed positions. One case of an intellectual act that Kant does explicitly discuss in relation to the exercise of the will is that of doxastic attitude formation or assent (i.e., “holding-for-true” or “Fürwahrhalten”).[46] Though it is somewhat controversial, Kant appears to straightforwardly deny that doxastic attitude formation is under one’s immediate voluntary control. For example, in the Jäsche Logik and related lecture notes Kant holds that the “will does not have any influence immediately on holding-to-be-true” (9:74; see also BL 24:156).

Kant’s conception of the “immediate” connection or influence of the will on assent is not thoroughly spelled out. As he explains it, the question is whether a phrase like “gladly believe”

Kant immediately goes on to deny this claim,

Kant denies the possibility of assent “at will,” which he construes as the immediate byproduct of choice, but allows that one can will oneself into picking up or dropping the investigation of the truth of some judgment, as well as the means by which such an investigation or inquiry is accomplished.[47] I take Kant’s denial here to, at the least, entail the denial that an act of will or exercise of the power of choice could be, by itself, the sufficient cause of one’s adopting, rejecting, or otherwise revising one’s assent to a judgment. In this sense Kant rejects what is now called “direct doxastic voluntarism.”

Kant’s rejection of direct doxastic voluntarism is important for answering the first objection in the following respect. Kant considers reflection on the grounds by which one assents to be, at least on some occasions if not generally, a “duty from which no one can escape” (Amphiboly, A263/B139; see also JL 9:76; BL 24:161). He also thinks that accepting something while deliberately avoiding further reflection is a kind of “punishable prejudice” (sträfliches Vorurtheil). It thus is plausible that Kant holds that assent to a judgment is an imputable act for which a subject can be held responsible, and for which she can reasonably be praised or blamed. Since Kant holds that assent is not, at least typically, something that is the immediate outcome of will or choice, Kant must not construe imputability as relying directly on one’s volition or power of choice.

What is much less clear is whether Kant construes imputability to require what he calls an “indirect” or “mediate” exercise of the will. If he doesn’t, then a lack of connection to volition is no problem for the general imputability of intellectual acts. But even if Kant does require that imputable acts have at least a mediate connection to the will, I think he can accommodate this via his conception of the centrality of attention for all rational activity.

When I introduced the notion of attention, in Section 2, as a kind of intellectual act, I argued that Kant understands attention as an act in which a subject can exert control over their consciousness of the various determinations of an object. This control over one’s representations comes from the subject in that it is independent of representational stimulus, cause, or occasion. Kant’s somewhat crass example is that of a suitor who “could make a good marriage if only he could overlook a wart on his beloved’s face, or a gap between her teeth” (An 7:131–32). The idea being, clearly, that the suitor would be happier if he were to exert attentional control over aspects of his representations, ignoring some elements while focusing on others. It is precisely this ability to exercise one’s will in controlling one’s attention that is lacking in non-rational animals. The attentional capacities of non-rational animals are determined in them by nature. The threats and promises that populate such an animal’s environment, along with its natural needs and inclinations, push and pull on its attentional resources, resulting in pathologically determined inclinations to act in various ways. This is why Kant construes the animal power of choice (arbitrium brutum) as “pathologically necessitated”, in contrast to the free (liberum) power manifested by rational beings (A533–34/B561– 62; see also Metaphysik L1 28:255 (c. 1778–1781); Metaphysik Mrongovius 29:896 (1782/3); MM 6:213).[48]

Hence, even if Kant regards free acts as being or involving the exercise of the will as ground of the power of choice, the central role of attention in acts of conceptualization, judgment, and reasoning allows the will a pivotal place in the determining grounds of a rational agent’s actions, even those that are not under its direct voluntary control, such as assent.[49] For this reason, I think the imputability argument might plausibly be extended to cover not just overt intellectual acts such as holding-for-true, but also acts of judgment, conceptualization, and perhaps other forms of synthesis as well.

Moving now to the second objection mentioned above, the worry is that Kant’s account of imputation derives entirely from his conception of practical or moral freedom, rather than from any feature of the intellect’s theoretical exercise. Hence, regardless of how we settle the issue of volition with respect to intellectual acts, the worry is that there are no specifically theoretical or cognitive norms. Instead all norms are inherently practical or moral. Patricia Kitcher puts the point this way,

If Kitcher is correct then there is no reason to think that imputability in the case of rational criticism of a thinker amounts to anything more than the moral requirement that the agent develop their (rational) talents (Kitcher 2011: 248).

However, Kant’s remarks in the Canon concerning assent indicate a more direct relation to laws governing imputability of assent than Kitcher’s interpretation allows. For example, he says that,

Kitcher might argue that the normative requirement in the first clause not to opine without knowing something relevant to the opinion’s truth is simply an expression of moral norms governing the development of proper reasoning skills. However, the rest of the passage indicates that the source of the normative requirement is the structure of assent as such. One cannot even count as opining (a form of assent) unless one’s opinion has a connection with the truth. Otherwise one merely plays a “game of the imagination.” This is not merely a condition related to developing one’s (rational) talents or skills. It is a requirement on assent.[50]

If these replies are cogent then the imputability argument provides at least some reason to think that any rational agent that is correctly open to criticism with respect to intellectual acts such as those of conceptualizing, judging, inferring, or assenting, must, by Kant’s lights, be transcendentally free. Note that the argument does not show that we are agents whose acts are imputable. The best the argument shows is the truth of the conditional claim that if agents are to engage in or correctly be objects of non-proleptic criticism or praise then they must exhibit control over their actions, and thus be transcendentally free. Moreover, the imputability argument does not provide a complete picture of Kant’s reasons for thinking that intellectual acts must be free. To get a better sense of Kant’s underlying view I examine below three further arguments for linking freedom and intellectual activity. The first concerns his conception of how reasoning works, which we’ll see requires that the intellect be transcendentally free in its activity.

4.2. The Reasoning Argument

Kant’s “mark” (Merkmal) theory of concepts construes concept possession in terms of recognition of conceptual content as the “ground” (Grund) of one’s cognition of a thing.[51] As he puts it,

Representation of a ground as such requires the generation of or transition to a mental state by recognizing the content of one’s present or previous state as the basis of the new state. Recognition of a ground of difference (or, relatedly, of similarity) is, for Kant, a recognition of a rational relationship between two (or more) things or properties. To take a simple example, one rationally distinguishes between two differently colored objects not merely in the sense that one’s visual system is sensitive to differences in color but rather because one also recognizes that one object is blue and the other red. In so recognizing this basis for distinguishing the objects one takes the color of the first object as a ground of its being different from the second object, and thus a reason for judging in one way rather than another. That is, in differentiating one object from another by their color one recognizes that, for example, blue objects cannot simultaneously be red and vice versa. Similarly, in thinking that an object is red one also thereby recognizes that its redness is reason to think it is colored and extended in space.[52]

It is crucial here that the subject not simply have states that flow one to the next in a manner that respects their content. It has to be the case that the subject appreciates these connections as such, and thus that her mental states are formed on the basis of such an appreciation of their content. Thus, the cognitive subject must play a causal role in the production of her states by recognizing that the content of one mental state connects with that of another and producing this other state on that basis.

Conceptual representation thus requires that the conceptualizing subject be able to base one state (or act) on another in the above manner. Kant distinguishes this kind of mental transition (call it “reasoning”)[53] from other kinds of mental transition (e.g., association in its various forms) in virtue of whether the subject has so based the succession of her states in a manner that recognizes the content of the one state as the basis for another. Following a variety of others, let’s call this condition on rationally connecting one’s states the “taking” condition, which says that a necessary condition of rationally basing one mental state (or act) on another—as happens in the inferential transition from one representational state to another—is that the subject take her state to be formed on the basis of her other state(s).[54]

If Kant does in fact endorse a taking condition on reasoning, and thus on conceptualization, then we can see how the variety of rational activities that he typically categorizes in terms of “synthesis” or “combination” (e.g., conceiving, judging, inferring, and assenting) are all going to be related via the central conception of basing one contentful state on another. We can then formulate an argument that such takings must be transcendentally free as follows.

1. Conceptual representation constitutively depends on the exercise of one’s capacity for reasoning—that is, the capacity for a particular kind of transition between one mental state and another.
2. Reasoning requires “taking” one state as the basis for the next.
3. Taking is an act under the subject’s control.
4. If the transition from one state to another is merely part of the “mechanism of nature”—that is, it is wholly determined by temporally preceding causes— then it is not under the subject’s control.
5. ∴ Mental transitions due to the mechanism of nature are incompatible with the subject’s engaging in reasoning—that is, reasoning requires transcendental freedom.

That there is a link in Kant’s view between “takings” and transcendental freedom has been, at least to some degree, recognized in the secondary literature.[55] What is less clear is why the act of taking is linked to such freedom. I’ve argued for why Kant accepts premises (1) and (2). We’ve also seen, in Section 3.2, that the temporally conditioned causal structure of nature is incompatible with the subject’s actions being under her control, and thus that Kant accepts (4).[56] The key then is to understand why he might accept (3). One reason might be that he accepts that we can hold people culpable for instances of poor reasoning, and thus that he relies on the imputability argument for justifying the acceptance of (3). But there are also considerations distinct from imputation, which stem from Kant’s conception of a faculty, that support his endorsement of (3). I discuss these in the next subsection.

4.3. The Constitutive Laws Argument

As discussed in Sections 2 and 3, Kant conceives of the various faculties and sub-faculties of the mind in terms of their causal activity. That is, for each faculty there is a characteristic set of effects for which it is responsible, and thus a characteristic set of acts that bring about these effects. Since causality is at issue in describing the relation between an act and its effect, the characterization of a faculty in terms of causal activity requires, for Kant, characterizing that faculty in terms of the constitutive causal laws governing that activity.

Because the laws governing the acts of a faculty are constitutive, the faculty, when not interfered with, always generates effects in accordance with those laws, for this is just what it is for the faculty to be what it is.[57] Kant is very clear about this point in the Transcendental Dialectic.

Kant construes error as the result of the deviation of a faculty’s activity from its characteristic laws in a kind of “geometry of error”, and which, since the issue is always one of causality, must be construed in terms of the causal interaction of different faculties upon one another.[59] As he puts it in the Vienna Logic:

Errors in thinking, for Kant, arise as a product of the influence of sensibility on the intellect.[60] So, the laws of the intellect produce particular effects unless interfered with by some external or “alien” causal power.

What are the laws by which the intellect operates? At the most general level they must include the laws of logic. Kant says,

These logical laws must be operative for the occurrences of and transitions between the mental states of a subject to count as genuine cases of thought or thinking. This means that in cases in which those laws are not operative, such as when the understanding’s activity is brought about through an alien cause, the status of these deviations is not characterized by Kant as thinking.[61] For example, in a note from a series discussing Meier’s conception of error in his Auszug Kant speaks of the “mixed effect” (vermischte Wirkung) generated by the influence of sensibility on the intellect, an effect that one would err in calling a “judgment of the understanding” (R2244, 16:284; see also JL 9:73).[62]

Kant also indicates in other published work that one’s rational states are generated according to the laws constitutive of the rational faculty’s operation. For example, in the Groundwork he famously says that,

Given Kant’s constitutive and causal conception of the activity of a faculty, it is understandable why he would say that reason cannot consciously “receive direction” (eine Lenkung empfinge) from something outside itself and still count as reasoning or judging, for what it is to reason, judge, or in general to think, is to generate specific kinds of mental effects in accordance with laws characteristic of one’s intellectual faculty rather than in accordance with laws as determined by something else distinct from that faculty.[63]

Hence, if the ultimate ground or cause of a mental state, or of the transition from one mental state to another, consisted in a causality originating in a power other than that of the intellect, the kind of “mentation” involved (if one can forgive the neologism) cannot be, for Kant, genuine thought. This means that, for thought to occur, the ultimate ground or cause of one’s thoughts cannot be due solely to the “mechanism of nature,” understood as the temporally structured causality of the natural phenomenal world. If one’s mentation were due to natural causal laws, then it would be ipso facto due to “alien influences” (fremde Einflüßen), and thus due to a causality distinct from that of the intellect.

We now have the material for the following argument, which makes these commitments explicit.

1. The acts of a faculty are governed by its constitutive laws.
2. The laws governing the intellectual faculty (der Verstand; Intellectus) are not those constituting the mechanism of nature.
3. If the ultimate ground of a transition from one mental state to another is due to the mechanism of nature then it would not be a transition determined by intellectual laws.
4. ∴ If a mental transition is intellectual/rational—that is, if it is to count as thinking—it must be both lawful and independent of the mechanism of nature and thus transcendentally free.

We can see now why, in the “taking argument” discussed previously, Kant must claim that taking (and thus reasoning) is something under the subject’s control. If taking were not under the subject’s control then an alien causality would be at work, and the resulting mental activity or state could not thereby be characterized as a case of thinking. The constitutive laws argument also helps us understand why the imputability of an act is so closely linked to the causal conditions of its production.

One might object here that intellectual acts cannot be transcendentally free in the way necessary for satisfaction of the constitutive laws argument, because they cannot plausibly constitute or be part of a causal chain initiated wholly by the subject. Patricia Kitcher puts the objection this way,

In reply it is important to see that this objection locates the initiation of a new causal chain in the wrong place.[64] Kitcher assumes that the acceptance of q stands at the beginning of a new causal sequence, and rightly worries that this would cut off the conclusion of an inference from its sufficient basis, invalidating the supposed rationality of the inference. But there is no need to interpret Kant’s claim in this way. It is not the acceptance of q, but a different mental act—the taking of p and if p then q as the basis for accepting q—that stands at the beginning of a new causal chain. This act is the act of relating premises to the acceptance of a new judgment, and if the ultimate causal ground of such an act is merely the mechanism of nature, or some other alien cause, then the result of that activity— viz. the transition to q—would not be an inference. It must be the causality of the intellect that is ultimately the ground of the inference to q and not merely that the preceding state of physical or psychological reality was thus-and-so.[65]

One might further object to the construal of a faculty’s operations in terms of causality on the grounds that it is implausible to understand logical laws as causal. In reply, I think we have to explore different notions of causality as they work in Kant’s system. The judgment that is the effect of an act of a faculty is, I think, supposed to be efficiently caused by the being (i.e., substance) whose faculty that is. As noted in Section 2, Kant construes faculties as the causal powers of substances to efficiently cause accidents, either in themselves or others. This is not a view that is unique to Kant, and goes back at least to Leibniz, who also holds that substances are the efficient causes of their states or accidents through an exercise of their causal powers.[66] But the fact that a particular judgment is efficiently caused by an act of judging does not rule out the possibility that Kant also thinks there are other kinds of causation necessarily involved in judgment. Indeed, in my view Kant also has to appeal to both formal and final causation. Let me briefly discuss final causation before turning to formal causation.

Final causation comes with what Kant calls the “interest” of a faculty:

Part of the ultimate explanation of a rational connection of one’s states concerns the fact that they are connected for a particular end determined by the fundamental interest of one’s rational capacities in their speculative use—viz. achieving cognition, and ultimately what Kant calls “comprehension” (Begreifen), via a priori principles (JL 9:64–65; see also Dohna-Wundlacken Logik 24:730–31 (c. 1792); Wiener Logik 24:846 (1780); Blomberg Logik 24:132–33, 134–35, 136 (c. 1771)). So, that for the sake of which all rational transitions involving the speculative (or theoretical as opposed to practical) use of our rational faculties are made is the greater understanding or making sense of things ultimately perfectly manifested in comprehension.

That the effects of a rational faculty have such a final cause says nothing about the structure in terms of which a determinate effect—e.g., a cognition—is efficiently brought about by the substance that has that faculty. It doesn’t, for example, say anything about the very different ways such an effect might be brought about by/in an intuitive intellect vs. a merely discursive one. But the appeal to the kind of structure necessary to so differentiate the two kinds of ways (viz. intuitive vs. discursive) of achieving the end of cognition looks to be just an appeal to the formal cause of the cognition. If this is correct, then Kant has to appeal to what looks like formal causation to give a full explanatory story of a determinate effect, like a judgment, being brought about according to intellectual laws (i.e., by rational causes). This would then mean that any rational act or transition in the mind has a formal cause, and that this would bottom out, at least for discursive minds, in Kant’s account of the unity of apperception, which is the intellectual form of all discursive cognition.

If this is all correct then while the intellect is that causal power of a substance through which it efficiently causes particular kinds of effects in accordance with laws of combination, a full explanation of the nature of this causal power and its effects will appeal to more than just efficient causation. Rather, such an explanation will also appeal to final causation (i.e., the aims of the cognitive discursive mind) and formal causation (i.e., pure apperception as the form of all intellectual cognition).

Kant’s position that the spontaneous exercise of one’s intellect works according to constitutive causal laws distinct from those of nature also helps shed light on one of his more perplexing arguments against hard determinism (or “fatalism”). The argument appears in Kant’s 1783 Review concerning Johann Heinrich Schulz’s work on ethics. There Kant says that,

Kant’s argument here is highly compressed. But one point he might be construed as making is that the proponent of fatalism undermines her own position because she can provide no guarantee that her judgments are based on reliable objective grounds rather than unreliable subjective ones, which as Kant points out “could subsequently change.” On this reading Kant thus argues that the fatalist lacks any justification for her holding true any of her judgments, for such judgments are just the lucky result of the deterministic order of nature. One problem with this reading of the argument is it seems to presume that assent to a judgment cannot be both naturally caused and based on rational grounds, but without further argument it is not clear why the fatalist should accept this.[67]

However, as I have presented Kant’s conception of a free intellectual faculty it is clear that Kant must find the fatalist’s attempt at argument incoherent. For it isn’t (or isn’t just) that if the fatalist is correct one’s judgments would be, at best, epistemically lucky (and thus unreliable). Rather, it is that according to the fatalist the judging subject would be, as Kant puts it, a mere “plaything” (ein Spiel sein) of her sensible impulses. In acting according to subjective grounds—that is, empirical natural (psychological) causes—she would not be acting in accordance with rational intellectual laws. Thus the fatalist would not be acting spontaneously but merely receptively and would thereby lack any imputability with regard to her acts. Hence, she could not be seen as engaging in anything like an argument concerning what is true. Kant holds that any argument the fatalist makes presumes a use of a free intellectual faculty—that is, a faculty whose causality is not simply that of temporally determined nature, and works instead entirely according to its own laws. The truth of fatalism would thus undermine any possibility of arguing for its truth, and in that sense any self-conscious attempt to argue for fatalism is self-defeating.[68]

There is a further issue that Kant’s conception of the constitutive causal laws of a faculty raises. What evidence do we have for thinking that we possess rational faculties that could be the ultimate causal sources of our mental states? Do we have any good reason to think that we, rather than the mechanism of nature, are the ultimate causal source of our thoughts? The constitutive laws argument, like the others above, tells us nothing about our status as rational beings; it merely indicates that rational beings would have to satisfy the conditions it outlines. In the next and final subsection I look at one argument Kant offers in favor of a substantive positive answer to the question of our status as rational beings. The view that emerges from his work, I argue, is one according to which we have at least some evidence for construing ourselves as genuine thinkers, but this evidence cannot ever rise to the level of proper objective cognition, and so cannot be construed as servicing a genuine form of positive scientific knowledge concerning our rational status as thinkers.

4.4. The Argument from Self-Consciousness

In the Transcendental Deduction Kant famously says that,

This text is typically interpreted as saying that any representation whose content could be something for the subject (i.e., present some possible, truth-evaluable state of affairs to the subject) must be such as to be self-ascribable. This is to read Kant as arguing only that rational subjects must be such as to possess the capacity for self-consciousness of the content of states that they undergo or to which they are otherwise subject. But Kant could plausibly mean something more in this passage than merely that in any act of thinking one must be able to become aware of oneself as a subject of that act.

In lecture notes Kant is reported as having argued for consciousness of the causal activity of the thinker in acts of thinking. For example,

Kant distinguishes here between two senses in which one might be self-conscious. One sense might be described as a kind of reporting—viz. the sense in which one is aware of oneself as a subject of representation, and thereby self-ascribes that representation. This is the sense that Kant is typically construed as having in mind in the Deduction passage quoted above. But a further notion used in the text is that of the subject as cause of the thought, or of thinker as the agent of thought. This “agential” sense is also at work in Kant’s reported rejection of Spinozism.

The point Kant is described as making in this passage concerns the manner in which the sort of self-consciousness characteristic of thinking—what Kant calls “pure” as opposed to “empirical” apperception—is a consciousness of oneself as the causal source of one’s thought, rather than of something else bringing that representation about in one. Kant also links here, as he does in the Analogies, causal activity and substantiality (A204/B250).[69] This link is specifically via the notion of being the ultimate causal progenitor of the thought. As Kant says above, “when I think, I am conscious that my I, and not some other thing” is what is doing the thinking.

In my view, Kant links the subjective and the agential sense of “I think” in a conditional manner. One cannot be non-empirically aware of oneself as the subject of a thought unless one is aware of oneself as the agent or cause of the thought.[70] One can of course be empirically conscious of mental states for which the agential condition is not satisfied; for example, in the empirical consciousness of inner and outer sense.[71] But what is important for cognition, as Kant notes in the “cogito” claim at B131–32, is the thinkability of one’s representations. Intuition is “given prior to all thinking” (B132), but if that intuition is to be something for the thinker (i.e., to play a role in cognition via its incorporation in experience) rather than to merely be in the thinker (and thus play at best a merely natural causal-mechanical role in determining the thinker’s actions), then it must be possible for the intuition to be taken up by the unity of apperception, and thus be a part of the intellectual activity of the cognizing subject.

I thus take Kant to have material for the following argument.

1. One cannot use <I> correctly in acts of conscious thought unless one has control of that thought.
2. One has control of a thought just in case the thought’s occurrence is due to the exercise of one’s intellectual capacities according to their constitutive laws.
3. The occurrence of a thought due to the exercise of an intellectual capacity is incompatible with its occurrence being due to the mechanism of nature.
4. ∴ One cannot correctly use <I> in thinking unless one’s thoughts occur due to intellectual causes distinct from the mechanism of nature—that is, unless one’s acts of thought are transcendentally free.

This argument hinges on the conceptual link between correct use of <I> and transcendental freedom. However, the claim that first person thought must be transcendentally free generates a clear interpretive problem. One might reasonably claim on the basis of this argument that at least from one’s own case one can know, or at least cognize, one’s transcendental freedom. This is because if the above argument is sound, then from one further assumption—viz. that anyone who can consider the truth of this argument is correctly able to use the first-person concept—knowledge of transcendental freedom is secured. Kant of course denies in the first Critique that we have any such knowledge of transcendental freedom (A558/B586). So mustn’t Kant reject the argument? I think Kant’s considered position is that the argument is sound. What he rejects is the further assumption of theoretical knowledge of one’s correct use of <I>.[72] We cannot have theoretical cognition of the fact that we are in control of our thoughts, so we cannot have theoretical cognition of whether we are in fact correct users of the first-person concept. Kant’s position on this point is not always consistently expressed.

For example, in the Clarification of the Cosmological Idea of Freedom Kant says,

The claim that one has cognition of one’s intellectual acts in pure apperception is rather problematic and we can get clear on why this is the case once we understand Kant’s conception of cognition. Kant uses the term “cognition” (Erkenntnis) in different ways.[73] For example, there is his very general definition of cognition as a conscious representation of an object (A320/B376). In this sense both intuitions and concepts may be cognitions. But even in this broad sense the claim that we cognize our free intellectual activity is problematic for there is no object given in pure apperception.

Kant also has a narrower sense of “cognition”—cognition in the “proper sense”—which, in finite beings such as ourselves, only concerns those states that are the outcome of a synthesis of sensory states with concepts.

I think we can see this narrower sense concerning the relation between concepts/judgment and intuition as motivated by a set of commitments that Kant endorses with respect to cognition—viz. (i) consciousness, (ii) agreement/conformability, (iii) real possibility.[74] Let me first briefly remark on each of these conditions before going on to show how pure self-consciousness fails to rise to the level of cognition proper.

The consciousness condition is clearly stated in Kant’s Stufenleiter. Cognition must be a conscious relation to an object. The conformability condition is also relatively straightforward. Kant conceives of cognition in terms of the “agreement” or “conformity” (Übereinstimmung) of a cognition with its object. For example, in discussing truth in the first Critique he says that the “nominal definition of truth, namely that it is the agreement of cognition with its object, is here granted and presupposed” (A58/B82). Since Kant allows for false cognition, I take his view to be that representations that are cognitions are such as to be able to agree or conform with their objects, and Kant construes such agreement as truth. So cognition need not be actually true to stand as cognition, though it must be such as to be possibly so. Moreover, Kant regards judgment (i.e., a specific sort of non-associative unity of concepts) as the bearer of truth, while he denies that intuition is the sort of thing that can be true or false.[75]

Finally, Kant construes cognition as always needing to be of a really—that is, metaphysically—possible subject matter. As Kant states in a famous footnote in the B-preface of the first Critique:

In finite discursive beings the structure of thought is governed ultimately by the rule of non-contradiction, and in this way concerns what is logically possible. But an analysis of the contents and laws of thought tells us nothing as to whether there really could be an object such as the thought specifies. Thus there can be merely discursively or logically possible represented objects (e.g., a two-sided plane figure) that are not really possible. There may also be really possible objects that are not representable discursively (e.g., the world-whole as a “synthetic universal”). Thus the real possibility of an object for theoretical cognition can only be determined from an actual ground provided via experience or via a priori reflection.[76] And as Kant goes on to note, the “more” that is required for a thought to achieve cognition need not always come from theoretical sources, as there may be practical sources of cognition as well.[77]

We can thus take “cognition in the proper sense,” at least in its theoretical guise, as necessarily (i) conscious, (ii) truth-apt (i.e., discursively conformable), and (iii) of a really possible subject-matter. Kant’s conception of cognition in such terms explains why he takes satisfaction of these conditions in finite rational beings (i.e., cognition in its “proper” sense) to require both intuition and concept, since discursive beings are such that they cannot satisfy all three conditions without the integrated contribution of both sorts of representation.

Kant’s critical notion of (theoretical) cognition in finite discursive beings thus requires a representational complex in which judgment and intuition are suitably related via an act of synthesis. The critical Kant is clear that one cannot have an intuition of the mental act by which thinking occurs.[78] Moreover, one cannot have cognition in the proper sense via pure apperception alone, since there is no (and could be no) corresponding intuition with which to synthesize the appropriate judgment concerning one’s intellectual activity. So I take it that Kant’s considered view must be that self-consciousness cannot ever provide cognition that one is the agent of one’s mental states. This also helps Kant avoid the objection that pure self-consciousness provides a sufficient basis for positively knowing that one is transcendentally free.[79] Such positive knowledge would require that one has cognition of one’s transcendentally free acts. But of such acts one can never have cognition in the proper sense; they are at best acts of which one is merely conscious.

We can see from this account, however, why Kant might be inclined in various places to nevertheless construe pure self-consciousness of one’s mental activity as a sort of cognition, for even in the second edition of the Paralogisms Kant maintains that “in the consciousness of myself in mere thinking I am the being itself” (B429) that is thinking, which is to say that in pure apperception I am non-empirically aware of my existence, and of the acts of thinking that I, as existing, bring about. This self-consciousness is both a form of consciousness and of an actual being (the thinker), and thus satisfies two of the three necessary conditions on cognition outlined above—viz. the consciousness constraint and the real possibility constraint.[80] What fails to be clearly satisfied is the conformity/truthaptness constraint, for one of the overall morals of the Paralogisms is that pure apperception does not provide sufficiently determinate material for reasoning about oneself, and thus nothing which one could articulate as a categorial claim concerning oneself as object.[81] Indeed, Kant goes so far as to say that,

Kant here denies that pure apperception presents any sort of sensory or intellectual appearance. Moreover, the “being itself” of which one is conscious in being engaged in intellectual activity is not itself represented in terms of an object. This is what I take Kant to mean in saying that the concepts <subject> and <ground> do not “signify the categories” <substance> and <cause>. I take Kant to be saying that one is aware in pure apperception of being the subject and ground of a thought, but that this awareness is not of the kind necessary to yield cognition, for the application of <subject> or <ground> here, and the related categories, is not warranted by any intuition. Thus there is a sense in which it is incorrect to construe the content of one’s thought here as constituted in terms of the application of <substance> or <cause>.

One might thus wonder in what sense the consciousness of oneself as thinker has any content at all, if it has no categorial content concerning a substance or its causal activity. Here I take Kant to construe the content of such self-consciousness as not including <substance> or <cause> even if it does contain the closely related <subject> and <ground> because of two reasons. First, the content of such self-consciousness is unschematized (it is not of a possible appearance). Second, lacking intuition of oneself in thought, there is no “object”—in Kant’s demanding and technical sense—presented to consciousness. Kant tends to use “content” (Inhalt) to mean “relation to an object,” and there is no such relation here. But this fact is compatible with our broader use of the term in contemporary philosophy, according to which “content” means something more like “information conveyed.”[82] Given that wider contemporary sense of “content,” Kant can consistently deny that pure self-consciousness provides any relation to an object, and thus material for judgment, without being forced to say that pure self-consciousness fails to convey anything at all to a subject.

Kant also acknowledges that how things seem in the consciousness of intellectual activity might not be the case. For example, while discussing the “Canon problem” in Section 3.1 we saw that Kant at least entertains the possibility that the activity of reason of which one is conscious might ultimately have its causal source in nature—in “higher and more remote operating causes” (A808/B831)—rather than the free faculty of reason itself.[83] So Kant allows that while it certainly seems to us as if we are engaging in rational acts such as inference, it is at least epistemically possible that we are not doing so. Thus we cannot know for certain that our reason is not ultimately determined by alien influences, and so cannot know for certain that we are genuinely thinking and reasoning, rather than engaging in their mere simulacra.[84]

However, despite our ultimate ignorance concerning our status as rational beings, Kant can also argue that we have sufficient grounds for the legitimate belief (Vernunftglaube) that we are such beings. Indeed, the claim that we are rational beings capable of being the agents behind our acts of attending, conceptualizing, thinking/judging, etc., is a necessary precondition or postulate of reasoning at all, on analogy to Kant’s conception of justified moral belief in the existence of God or of our own immortal souls.[85] This sort of “doctrinal belief”, as Kant calls it (A825– 26/B853–54), is a theoretical analogue of practical belief, and concerns that for which we can never have sufficient objective grounds for holding true, but which is something to which we are nevertheless rationally driven.[86] Moreover, this drive is distinctively theoretical rather than practical, concerning the interest of reason in what is rather than what ought to be (CPrR 5:120; A804/B832).[87] Following Stang (2016a: 263) we can call the kind of belief at issue here “necessary theoretical belief,” which is belief in a postulate that is necessary to satisfy theoretical reason’s need for explanatory unity, but whose truth cannot be cognized or known by us. As Stang (2016a: 287) points out, such necessary theoretical belief, if it is to be relevantly analogous to moral belief, must be of something that is knowably logically possible and must not knowably be really impossible.[88] Our status as rational causes of our intellectual acts satisfies this, at least insofar as the argument that we can be progenitors of new causal chains, as set out in the Third Antinomy, is cogent.[89]

Consciousness of our representational activity via pure apperception thus provides at least some reason for thinking, as the constitutive argument demands, that we have rational faculties and thus could be the ultimate causal sources of our intellectual acts. But we can never, at least on theoretical merits, know or even cognize that we have such faculties, and thus know or cognize that we engage in such transcendentally free acts. Nevertheless—in analogy with the moral case—it seems that we have a legitimate basis for belief (Glaube) in our intellectual freedom.

5.1. Relative Spontaneity

Perhaps the most well-known objection to the position that Kant conceives of the intellect as transcendentally free is from Wilfred Sellars, who argues that,

And slightly later,

Sellars is arguing here for two points. First, that Kant distinguishes between an “absolute” spontaneity of transcendental freedom and a merely “relative” spontaneity, that is appropriate only to a “mechanism.” Second, that Kant allows for the possibility that the theoretical use of the intellect in the cognition of objects is merely relatively spontaneous.

We’ve seen, in Section 3.2 above, that Kant rejects the kind of temporally determined relative spontaneity proffered by a broadly Leibnizian model. Let’s assume that Sellars doesn’t have this kind of position in mind. But then what exactly is the sense in which the theoretical use of reason evinces a spontaneity that is merely “relative,” such that it doesn’t allow for imputation, in contrast with the practical use of reason, which does?

The main problem, as Sellars conceives of it, is that the theoretical use of one’s intellectual faculties might work according to principles, while nevertheless lacking any purpose or norm-providing end towards which one’s acts strive.[91] The way that Sellars puts this difference between the theoretical and practical uses of reason is in terms of the presence in the latter use of what he calls a “practical premise”:

The implication here is that the theoretical use of reason is not governed by a “premise” intrinsic to it, rendering it merely relatively spontaneous. The “premise” is supposed to be something more than a simple “transformational” rule or principle, rather it must be something that articulates a purposeful use of or end for the activity of the faculty. Sellars ultimately attributes such purposeful activity only to the practical intellect. He says,

Sellars argues that while, for Kant, self-love is a purpose “implanted in practical reason from without” the moral law counts as an internal purpose or “intrinsic practical premise” (1971: 27). Interpreted this way, Sellars argues that the practical use of one’s faculties involves an intrinsic purpose for their use that is not present in the theoretical use of one’s faculties, and for that reason the theoretical use of one’s intellectual faculties is merely transformational (i.e., “relative”) and not genuinely (or “absolutely”) spontaneous.

But positing such a disparity isn’t cogent. Most importantly, Sellars draws a thoroughly false contrast between the merely “transformational” procedural acts of the theoretical use of reason and the “intrinsically purposeful” acts of practical reason.

As I discussed in Section 4.3, Kant distinguishes the two uses of reason according to the distinct intrinsic aims that their use accomplishes, each of which cannot be reduced to the other.

The interest of theoretical reason consists in cognition of what is, in accordance with “the highest” of a priori principles provided through the unifying and systematizing activity of theoretical reason. The interest of practical reason lies in the determination of the will, again from a priori principles, concerning what ought to be, in the sense of realizing the “final and complete end” of the highest good. But because each use of the intellect has an intrinsic aim or end, exercises of that faculty are themselves governed by that end. For example, one cannot engage in acts according to the interests of theoretical reason if those acts are not in accord with laws that are or are suitably related to logical laws.[92] Sellars is thus wrong to conflate the interest of reason in its “practical aspect” with the interests of reason simpliciter, and he is wrong to think there is a special disparity to be found between the intrinsic purposes of the practical as opposed to theoretical use of reason (or the intellect more broadly).

If Sellars’s argument as stated cannot drive a wedge between the theoretical and practical uses of reason (or the intellect more broadly) based on appeal to purposes, might there not be other ways? Might the theoretical use need “material from without” in a way that the practical does not? Or might the intrinsic principles governing the practical be more clearly specifiable than those governing the theoretical? Or, finally, might the practical use of reason be “spontaneous” in the sense of providing material for reasoning that is novel by going beyond any possible sensory input? I believe Kant can give straightforwardly negative answers to all three of these attempts to differentiate the spontaneity of the practical from the theoretical use of reason.

First, practical reason, just as with its theoretical use, requires material upon which to work, so there is no disparity between their characteristic forms of activity to be found at the point of what initially causes reason to spring into action—in both cases some input from sensibility is necessary.[93] Hence, if we take Sellars’s initial gloss of a relative spontaneity at face value, as that which is “set in motion by foreign causes” it is unclear that there is any difference between the practical and the theoretical elements of the intellect. All of one’s intellectual faculties ultimately depend on being first set in motion by the deliverances of sensibility.[94]

Second, as we saw in the discussion of the constitutive laws argument, the intellect must have principles intrinsic to its activity in order for it to be individuated as a distinctive faculty or power for the generation of representations. Such principles presumably include the principle of contradiction, which Kant says is the “supreme principle of all analytic judgments” (A150/B189). Since general logic concerns analytic relations (e.g., A76/B102) and the “absolutely necessary rules of thinking, without which no use of the understanding takes place” (A52/B76), the principle of contradiction is clearly an intrinsic principle governing the operation of the intellect as such. It seems plausible that similar considerations can be raised for the operation of reason and the form of the principle of sufficient reason Kant articulates as the principle driving reason’s search for the conditions of any given conditioned thing (A307/B364).[95] So the theoretical use of the intellect has clearly articulable intrinsic principles and premises just as the practical use does.

Third, it is part of Kant’s conception of theoretical reason (here in its narrow sense as faculty of inference and progenitor of ideas) that its activity is not exhaustively understood as a response to sensible material—“it does not follow the order of things as they are presented in intuition, but with complete spontaneity it makes its own order according to ideas” (A548/B576; see also GIII 4:452). Theoretical reason is thus not merely reactive to sensory stimuli in activities of reflection and synthesis, as is the understanding, but rather generates its own representations—ideas—by virtue of the principles governing its activity (specifically, the aforementioned principle of sufficient reason governing its search for systematicity and unity).

Sellars might concede that theoretical reason has its own constitutive principles, and that its actions outstrip what is given to it by sense so that it is not a merely reactive faculty, but nevertheless deny that possession of such constitutive principles of activity is sufficient for the kind of cognitive control and absolute spontaneity for which I have been arguing.[96] Sellars suggests that the relative spontaneity of the theoretical use of the intellect might be like that of a computer realizing “set dispositions” (1971: 25) in response to an external stimulus. I’ve advanced a conception of “cognitive control” in terms of acting according to the constitutive laws of one’s cognitive faculties without being determined to do so by an alien ground or cause. Isn’t this just action in accordance with “set dispositions,” as Sellars describes? If so then how is it an act of absolute spontaneity—that is, transcendental freedom?

It cannot be action from “set dispositions” per se that is the problem here, for God’s intellect works from set dispositions—laws from which he never deviates— and this fact does not show that God lacks absolute spontaneity. Moreover, one can object here to the very notion of describing intellectual capacities or abilities as “set dispositions” since <capacity> and <disposition> are not identical concepts. One can have a capacity without having any specific set disposition— that is, I have the ability or capacity to throw myself out a window, but I lack any disposition (settled or not) to do so.[97] In any case, the crucial idea is that intellectual acts occur according to specific, constitutive, laws. If Sellars is insisting here that an absolute spontaneity doesn’t operate according to set laws but rather must be able to deviate from them, then perhaps he is making one of two illicit assumptions.

On the one hand, Sellars might be operating with a conception of absolute spontaneity as (or as requiring) leeway, which I have already argued (in Section 3) that Kant rejected. Of course, there are ways in which one might exercise choice with respect to intellectual acts—one can choose what to attend to, what lines of inquiry to adopt or pursue, etc.—but one is not acting according to one’s rational powers if one refuses to accept q despite accepting p and if p then q, or if one tries to think a contradictory thought of the form (p & ¬ p). In lacking the ability to do other than think a logically coherent thought or make an obvious logical inference based on premises to which one both attends and accepts one does not thereby forsake one’s freedom/absolute spontaneity, rather one exemplifies that very freedom.

On the other hand, Sellars might instead assume that the causality with which one acts need not always be lawlike, but Kant clearly denies this.[98] And if Sellars’s issue ultimately concerns neither of these points but rather that he holds that the theoretical use of the intellect is merely responsive in a manner that contrasts with the activity of its practical use, then we fall back on the various sorts of response I have given above.

Hence, pace Sellars, the theoretical use of the understanding and reason, just as with the practical use of reason, depends on the existence of principles constitutive of the causal activity of the faculty. Since these principles are rationally constitutive they will not count as “alien” influences stemming from sensibility or implanted by God.[99]

There is one further issue about relative spontaneity, viz. that what “haunts” Kant is the possibility that we are merely spiritual automatons (Sellars 1971: 25). Recall from Section 4.4 Kant’s argument in the Metaphysik L1 and Metaphysik Pölitz lectures that one cannot be non-empirically aware of oneself as the subject of a thought unless one is aware of oneself as the agent of the thought. Sellars’s “haunting” question might then be put in terms of whether one could conceive of oneself as the merely relative agent of one’s thought, so that all of one’s self-conscious activity could ultimately be explained by mechanical causes beyond one’s control.

Consider a mundane example. While talking with a friend, one has the spontaneous mental image of one’s keys on the table at home (the image occurs to one “in a flash”), which then is the basis for one’s judging that one left one’s keys at home. This might seem an instance of “relative spontaneity” in one’s thinking, governed by certain conditions from without, but proceeding according to internal rules. Why couldn’t all thought be described in this manner?

While such episodes are colloquially described as episodes of “thinking,” Kant would consider that description contentious. At most, he would argue, the occurrence of the mental image itself is merely “relatively” spontaneous in the sense that it is generated by the imagination in response to stimuli. In contrast, the thought that one’s keys are at home, insofar as it is a rational judgment based on the content of the image, and not itself a mere happening, is something over which one has cognitive control. It is thus an act that one actively does, and in principle could refrain from doing (e.g., as when one knows one has been given a drug that will prompt the occurrence of random visual images). If the resulting state/activity isn’t a rational response to the image, but itself a kind of mere happening, Kant would deny that it is a judgment at all, being instead a kind of sensory association (as when one “hears” oneself “say” in an episode of internal speech, that one’s keys are at home).[100] In this latter case, what we would have is, at best, mere association, and not spontaneous thinking in the Kantian sense at issue. Moreover, the kind of mentation (to use a neutral word) involved in such mundane episodes as the example above describes is notably different in character from the experience of marshaling one’s thoughts in judgment or inference.

Thus Sellars’s view that thinking might be merely relatively spontaneous is simply contrary to Kant’s account of thought. If this is the possibility Sellars is suggesting bothers Kant, he would be wrong to allow that Kant is haunted by, or otherwise might endorse, such a view of the intellect. There is a related worry bothering Kant, but Sellars fails to adequately articulate it. What haunts Kant is the possibility that we may not be rational beings, our mental activity at best being a kind of simulacrum of a rational mind. He is thus not haunted by the possibility that we may be merely relatively spontaneous thinkers while being absolutely spontaneous actors, for there is no principled basis in Kant’s view for the relevant contrast. Instead the issue is whether we are genuine or merely ersatz rational agents. In Section 4.4 I argued that Kant has the resources to claim that we are legitimately driven to our belief that we are rational, and thus that we are genuine thinkers, though this belief can never be given a proof on purely theoretical grounds. But the principled basis by which Kant can argue for the legitimacy of the belief should be enough to exorcise the specter of Sellars’s “haunting” possibility.

Hence, in my view Sellars fails to provide for a substantive distinction between the form of spontaneity operative in practical as opposed to theoretical uses of the intellect as Kant conceives of it.[101]

5.2. Kant’s Change of Strategy

As noted at the outset of this essay and often discussed in the secondary literature, Kant pursues a strategy for demonstrating our transcendental freedom in the early to mid-1780s that links together theoretical and practical spontaneity. This is reflected in Kant’s 1783 Review concerning Johann Heinrich Schulz’s work on ethics, discussed in Section 4.3. Kant’s strategy for proving one’s practical freedom through an appeal to the freedom of the intellect is taken to culminate in the Groundwork III argument for the autonomy of practical reason.[102] However, Kant subsequently appears to discard this strategy, and perhaps even the entire aim of providing a demonstration of freedom, opting instead for the infamous “fact of reason” argument in the second Critique.[103]

This raises the following worry: perhaps at one time (e.g., the 1770s) Kant accepted that the intellect is transcendentally free, and pursued a strategy consonant with that position. But at least by the late 1780s his rejection of the strategy linking the theoretical and practical uses of reason might indicate that he no longer endorses the view that the intellect is free, at least in its theoretical use.[104] If that were correct, then either Kant’s overall view is incoherent, or I am wrong about the conceptual connections between the nature of the intellectual faculty and freedom.

Against this worry, I argue that Kant maintains his view concerning the free intellect throughout the critical period, but subsequent to the Groundwork holds that cognition concerning the free activity of one’s faculties can only be supported by practical as opposed to theoretical grounds. Hence, while Kant’s conception of the epistemic grounds for accepting that we have such a free intellectual faculty evolves over the course of the critical period his basic conception of the nature of that faculty does not change. There are three considerations for favoring my preferred interpretation.

First, showing that the activity of the intellect even in its theoretical use, and not just its moral or practical use, presupposes transcendental freedom does not in and of itself provide us with overwhelming justification for holding that we can behave morally. The argument of Groundwork III in part consists of a “partners in crime” style argument that reason in its production of ideas necessary for theoretical cognition, exhibits a form of spontaneity that transcends all empirical conditioning. But a committed skeptical naturalist might deny this claim about theoretical reason just as thoroughly as any claim about the freedom of practical reason. It’s not clear then, that Kant’s argument in Groundwork III provides any real traction against a committed skeptical naturalist.[105] Indeed, acknowledgment of this fact all on its own, without requiring a change to Kant’s underlying commitments, could have prompted him to change his approach in the second Critique, where his appeal to a “fact of reason” seems to be deliberately indifferent to skeptical worries.

Second, even if one accepts that the use of the intellect requires the transcendental freedom of that faculty, this does not show that the freedom of the faculty is such as to be able to determine the subject’s power of choice (Willkür). For it is one thing to argue that a faculty can determine itself, it is another thing to argue that it can determine another faculty—that a rational law could be a law that governs choice.[106] There is no contradiction in conceiving of a being possessed of an intellectual faculty whose operations satisfy the conditions for transcendental freedom, but which lacks the power to determine its faculty of choice according to reason (e.g., perhaps because its choices can only be sensibly determined, or are wholly lawless). This point becomes especially salient once Kant clearly differentiates between will (practical reason; Wille) and choice (Willkür).[107]

Third, since, as I argued in Section 4.4, pure self-consciousness of one’s intellectual activity can never rise to cognition proper, any attempt to substantiate (theoretical) cognition or knowledge of one’s freedom is going to fail. This means that there can be no demonstration of one’s freedom based solely on premises appealing to theoretical grounds.[108]

Thus, insofar as Kant’s views in the 70s and the early-to-mid-80s concerning the extent of our theoretical cognition are in flux, he is thereby drawn to a view according to which the conscious awareness in pure apperception that one has of the self and its activity is enough for cognition (e.g., A546–47/B574–75, discussed above). But by the second Critique Kant more consistently recognizes that consciousness alone cannot provide for the cognition necessary to the pursuit of a demonstration of our freedom.[109] This more consistent position would not require Kant to change his mind concerning the connection between the intellect and transcendental freedom. But it would affect his pursuit of any argumentative strategy concerning the aim of demonstrating our transcendental freedom.

Given the above three points, it is perfectly consistent for Kant to continue to hold in the mid-to-late critical period that the intellect is transcendentally free while coming to see that our status as rational agents cannot ultimately be proven on theoretical grounds, and that a shift to a proof on practical grounds is more consistent with the structure of the critical system as a whole. In this way the four arguments we’ve considered in Section 4 all outline the contours of a theory of rational activity without, on theoretical grounds, establishing that the human being satisfies such conditions. It is notable though that there do not seem to be considerations for showing that the human being cannot satisfy such conditions either.