HENRY E. ALLISON | Kant’s Transcendental Deduction. An Analytical-Historical Commentary | Oxford: Oxford University Press, 2015


By Henry Allison

I wish to thank Lucy Allais both for her extremely generous comments and her attention to many of the often messy details of my account. She has given my book what I believe any self-respecting author should want: a close and fair minded read. I am particularly pleased by the many points of agreement that she notes and by the fact, which she also notes, that many of our disagreements are relatively minor. Since she had taken a somewhat more combative stance towards my non-metaphysical or, as she put it in her book, “deflationary” interpretation of this idealism in her own recent book (Allais 2015), I was pleasantly surprised to see that even in this area we share more common ground than I had previously supposed. For while I take a stronger stance by claiming flat out that the Deduction is inseparably intertwined with transcendental idealism, while she maintains that significant elements of Kant’s argument do not rely on this idealism, she also notes that for Kant the role of the categories in constituting experience is epistemological rather than metaphysical, which accords with my understanding of Kant’s idealism. Moreover, she evidently acknowledges that some idealistic commitment is implicit in the claim that our cognition is limited to appearances, since, as she puts it, “[a]t the very least, ‘appearance’ here must mean things that are presented to us”, which I take it means that she acknowledges that the Deduction presupposes the Aesthetic.

Since Allais notes three background assumptions of my approach to the Deduction that have both significant similarities to and differences from hers, I shall organise my response around these and attempt to address some of her specific criticisms within this framework. The first and the one that I shall have the least to say about is transcendental idealism, because, as noted above, our differences do not seem to be that great, at least not according to what she says in her comments about my book. Nevertheless, it is clear that there is one fundamental area of disagreement regarding the relation between Kant’s idealism and the argument of the Deduction. Allais suggests that I consider Kant to be there arguing for this idealism; whereas in my view he is arguing from it. Kant did not need to argue for transcendental idealism in the Deduction, because he had already done so indirectly in the Aesthetic, where it was shown to follow from the ideality of space and time; and he argues for it again, albeit indirectly, in the chapter on the Antinomies. I maintain that the Deduction argues from this idealism for two reasons. First, as Allais acknowledges, it presupposes the results of the Aesthetic, which means that the conditions of cognition that it investigates are conditions of the cognitions of objects as they are given to us through our forms of sensibility, that is, conditions of things as they appear, not as they may be in themselves. Second, the argument proceeds from what I call a transcendentally internalistic standpoint. By internalism I understand the epistemological position according to which knowledge claims are analysed and justified in terms of the cognising subject, that is, by considering what counts as cognition for such a subject. In the book, I contrasted Kant’s internalism with that of Tetens, who exercised considerable influence on Kant at the time of the composition of the A-Deduction. And whereas I characterised Tetens’ internalism as psychological, because he was concerned with the mental processes through which human cognisers construct for themselves representations of objects that are assumed to have a real existence independently of these representations, I deemed Kant’s version transcendental because of its focus on the subjective conditions through which such cognisers represent to themselves an objective world (pp. 148–50). Moreover, because of this focus I consider Kant’s stance to be idealistic in a broad, methodological sense, which is why I often refer to his idealism as methodological or epistemological in nature, in order to distinguish it from a metaphysical view such as Berkeley’s.

The second background assumption is the distinction between sensibility and understanding as two generically distinct cognitive powers, each with its own set of a priori forms. Although Allais acknowledges the importance of this distinction for the Deduction, she seems to understand it in a somewhat different way than I do. According to her, it apparently lies in the fact that some commentators “sometimes see Kant’s concern in the Deduction as being with showing the categories to be conditions of objects being given in intuition”, which, she suggests, “results in running together the concerns of the Deduction and the Aesthetic […]”. Since Allais does not cite any examples of commentators who are guilty of this error, I do not know whom she has in mind; though I suspect that they must be what she refers to as “strong conceptualists” (Allais 2015:149). But while I share her rejection of such interpretations, inasmuch as they stem from what I consider to be a gross misreading of the text, for that reason they do not strike me as a major concern.  

By contrast, my view is that the importance of this distinction underlies the problematic of the Deduction, which I take to be the worry that what is given in sensibility might not accord with the requirements of the understanding, which I characterise as a spectre that the Deduction endeavours to exorcise. Allais is well aware of this, but she questions whether this worry “follows directly from the separation of the cognitive roles of sensibility and understanding”. I think that she is correct in raising this question, since it is clear from Kant’s position in the Inaugural Dissertation, where sensibility is considered the faculty for cognising the sensible world and the intellect the intelligible world (Kant had at that time not yet drawn a sharp distinction between the understanding and reason) that at least at that time he did not. Moreover, I did not claim that he did. My view is rather that this separation is a necessary, but not a sufficient condition of the worry underlying the problematic of the Deduction. It is obviously necessary because it underlies the possibility of a lack of a cognitive fit between concepts and intuitions as species of representation, which, it should be noted, could not arise for either the Leibnizians, for whom sensible representations are obscure versions of what is grasped clearly and distinctly by the intellect, or for Hume, since he considered ideas as merely pale copies of impressions. It is not sufficient because this worry also assumes what I term the discursivity thesis, that is, the view that cognition requires that what is given in intuition conforms to what is thought by the understanding, which I show came later in the development of Kant’s thought than the sensibility-understanding distinction.

The third assumption that Allais notes is that the main target of the Deduction is ‘concept empiricism’, that is, the claim that all of the materials of cognition are derivable from experience, and that there is, therefore, no need for a priori concepts in order to account for the possibility of empirical cognition, rather than a Cartesian-type scepticism regarding the reality or objectivity of experience, since that is the task of the Refutation of Idealism. Moreover, in agreement with Karl Ameriks (1978), I conclude from this that the Deduction is best read as a regressive argument, which assumes that we have experience in the sense of objectively valid empirical cognition and endeavours to show that a set of a priori concepts, namely, the pure concepts of the understanding or categories are necessary conditions of its possibility. And while I am not sure how Allais stands on this issue, I believe that she would likely concur, because an argument that a concept empiricism is incapable of accounting for empirical cognition would seem to have little point, unless it is assumed that we have such cognition and that the worry concerns the conditions of its possibility.

Nevertheless, Allais evidently was not completely satisfied with the arguments that I provided on Kant’s behalf for this admittedly sweeping claim. And since her dissatisfaction seems to have been directed mainly at two closely related points I shall focus on these. The first and most important is what I have termed Kant’s non-contingency thesis, that is, his seemingly paradoxical claim that empirical cognition involves a peculiar kind of necessity that is neither causal nor logical, and that for lack of a better name I have dubbed normative. Like Laywine, who made it the focal point of her reply, Allais expresses puzzlement about this, and I am afraid that much of what I have to say on the matter will involve repetition of points made in response to Laywine. I shall begin with the examination of the following passage, since it contains virtually all the points that are in contention and a good portion of Allais’ dissatisfaction seems to concern what I say about it:

Unity of synthesis in accordance with empirical concepts would be entirely contingent, and, were it not grounded on a transcendental ground of unity, it would be possible for a swarm of appearances to fill up our soul without experience ever being able to arise from it. But in that case all relation of cognition to objects would also disappear, since the appearances would lack connection in accordance with universal and necessary laws, and would thus be intuition without thought, but never cognition, and would therefore be as good as nothing for us. (A111)

The passage as a whole contains a reductio of concept empiricism. It attempts to show that the view is incoherent because it implicitly denies the conditions of the possibility of forming and using the very concepts that it claims to be necessary for empirical cognition. In effect, it maintains the seemingly paradoxical claim that if there were only empirical concepts, there would be no genuine concepts at all, but merely “intuition without thought”; and inasmuch as cognition requires concepts as well as intuitions, this would lead to the denial of the possibility of empirical cognition and, therefore, of experience. And the reason proffered for this is that on this scenario the unity of the synthesis, which is the synthetic unity that constitutes an empirical concept, would be “entirely contingent”. Moreover, this claim raises at least three questions, which get to the heart of both Kant’s procedure in the Deduction and some of Allais’ qualms about my account of it:

(1) Why did Kant claim that, absent some kind of transcendental ground, understood as an unspecified non-empirical contribution, the synthetic unity that constitutes an empirical concept would be contingent? 

(2) Why did he think that this contingency would yield such disastrous consequences and how did he construe the alternative?

(3) What did he mean by a transcendental ground of unity and how is it supposed to remedy the situation? 

(1) Addressing the first question requires a consideration of Kant’s conception of concepts in general and empirical concepts in particular. Although it is well known that concepts for Kant function as rules, for understanding the intricacies of Kant’s position it is essential to keep in mind that he also considered them as synthetic unities. To reiterate what I said in my reply to Laywine, concepts for Kant are synthetic unities in the sense that they contain a many, i.e. the partial representations or ‘marks’, which are themselves concepts, in one thought. Accordingly, the included concepts constitute the intension or what is thought in the concept. Thus with the concept of body, which is Kant’s paradigmatic empirical concept, we think of something that is extended, impenetrable, has a certain shape, etc. As was also noted in my reply to Laywine, however, the situation is complicated by the fact that Kant considered concepts not only as synthetic unities, but as analytic unities as well. They are the latter in the sense that they contain a many under one, e.g. the various bodies that fall under the concept of body. Moreover, both kinds of unity are essential to Kant’s complex conception of concepts.

Our present concern, however, is with concepts as synthetic unities and its bearing on Kant’s worry about contingency. Here the crucial point is that, as such unities, concepts for Kant are products of the understanding and are produced through what he characterises as a complex “logical act” of comparison, reflection, and abstraction (Log, AA 9:95). Simply put, concepts are constructed or made by the understanding, and this brings with it a worry about the possible arbitrariness or contingency of the process through which they are made; while this, in turn, calls into question their adequacy as vehicles of cognition, or, in Kantian terms, as predicates of possible judgements. I shall attempt to clarify what I take to have been at stake for Kant with regard to the element of contingency in the formation of empirical concepts by briefly contrasting it with the view of Locke, who was the quintessential concept empiricist; and, in light of this, I shall call attention to the subsequent deconstruction of this view by Hume. Like Kantian empirical concepts, Locke’s complex ideas are products of what he termed the “workmanship of the understanding”, through a combinatory process in which the understanding “puts together several of those simple ones [ideas] that it has received from Sensation and Reflection, and combines them into complex ones”. But particularly germane to our present concerns is his account of the genesis of the ideas of particular substances, which he claims are combinations of simple ideas that “are taken to represent discrete particular things subsisting by themselves; in which the supposed or confused Idea of Substance, such as it is, is always the first and chief” (Locke 1975:165). In his chapter dealing with our complex ideas of substance, Locke states that 

[t]he Mind being […] furnished with a great number of the simple Ideas, conveyed in by the Senses […] or by Reflection on its own Operations, takes notice also, that a certain number of the simple Ideas go constantly together; which being presumed to belong to one thing […] are called so united in one subject […] [and] not imagining how these simple Ideas can subsist by themselves, we accustom ourselves, to suppose some Substratum, wherein they do subsist, and from they do result, which therefore we call Substance (Locke 1975:295).

The language of this passage reflects Locke’s deep ambivalence about the idea of substance. On the one hand, he recognised that it cannot be derived from either sensation or reflection, and since these are the only sources of ideas according to his empiricistic programme, this makes its very existence problematic for him. Thus, at one point he denied that we have a proper idea of substance at all (Locke 1975:95); while elsewhere he famously characterised it as referring to “something, he knew not what” (1975:296). At the same time, however, Locke evidently thought that he could not dismiss the idea of substance completely because he considered it necessary to provide a substratum for the various qualities (both primary and secondary) that are predicated of objects in a judgement.  

Nevertheless, our present concern is not with Locke’s tortured endeavour to incorporate the concept of substance into his empiricistic project; rather, it is with his account of the formation of the ideas of substance-types or, as he terms them, “specifick Ideas of Substances”, which he claims “are nothing else but a Collection of simple Ideas, considered as united in one thing” (Locke 1975:305). Although it seems clear that for Locke the idea of substance is supposed to do the unifying, a view with which Kant would certainly accord, the issue is the ground for this unification or, as Locke termed it, “Collection”. Locke’s answer, as gleaned from the text, appears to be that the component simple ideas “go constantly together”. Expressed in Humean terms, these simple ideas, which Hume termed impressions, are joined together either by contiguity or constant conjunction (with the latter being the basis for the relation between cause and effect). Moreover, since for Hume these unifying bonds are principles of association governing the imagination, this joining together, or, to use the Kantian term, ‘synthesis’, is seen to be the product of the imagination, rather than of the workmanship of the understanding, as it presumably was for Locke. And that is why I suggested that Hume might be seen (and I believe was seen by Kant) as deconstructing Locke’s workmanship of the understanding at a significant point.

Moreover, I also take this to explain why Kant thought that “unity of synthesis in accordance with empirical concepts would be entirely contingent”. The contingency at issue concerns the agreement between the order of representations (Lockean simple ideas, Humean impressions, and Kantian empirical intuitions) in the mind and the order of nature, or, simply put, between the subjective and the objective orders. Since their agreement is considered by all parties as a necessary condition of the cognition of the latter order, this contingency calls into question its cognisability. Hume brought this issue to centre stage, when, in a passage that must have caught Kant’s attention, he refers ironically to “a kind of pre-established harmony between the course of nature and the succession of our ideas” (Hume 1999:129). And I further take Kant to be responding to this, when, in a discussion of the importance of Hume for the development of the ‘critical philosophy’ in the Transcendental Doctrine of Method, he remarks that, Hume “made a principle of affinity, which has its seat in the understanding and asserts necessary connection, into a rule of association, which is found merely in the imitative imagination and which can present only contingent combinations, not objective ones at all” (A766–7/B794–5). While this is presented as a criticism of Hume, it also credits him for highlighting the problem that became the central concern of the Deduction.

(2) We have already seen that for Kant the consequences of a pure empiricism, understood as one that does not admit any a priori components in its account of empirical cognition, such as Locke had tried but failed to provide, are dire; since they constitute nothing less than the denial of the possibility of the concepts on which experience as empirical cognition is based. But as any student of the Deduction is well aware, the reasoning underlying the view is both extremely complex and controversial. Accordingly, after devoting a lengthy book to the topic, which has at some points elicited Allais’ puzzlement, I can attempt to do no mere here than to rehash the main point, which, baldly stated, I take to be the claim that the kind of contingency at issue is incompatible with the objective validity of empirical cognition, because it undermines the peculiar species of necessity that Kant also claims such cognition requires. 

Although Kant expresses the point in a number of ways in both versions of the Deduction, I find the most perspicacious account to be the formulation in the Prolegomena. Moreover, I say this despite the fact that the distinction that Kant there draws between judgements of perception and judgements of experience, with objective validity and a role for the categories attributed to the former, and merely a subjective validity to the latter, appears to contradict the account in the B-Deduction, where Kant considers judgement as such to possess objective validity (§19) and connects this with the categories (§20).[1] Simply put, a judgement of perception has the form: “It seems to me that x”; or, to cite some of Kant’s examples, “The room is warm, the sugar sweet, the wormwood repugnant” (Prol, AA 4:299). Such judgements are first-person reports concerning how things seem to a particular perceiver at a particular time. Accordingly, their common feature may be characterised as a two-fold non-projectibility: to other perceiving subjects and to other times, which Kant takes to mean that they have merely a subjective validity. Nevertheless, Kant characterises them as judgements, rather than mere associations, because they contain “the logical connection of representations in a thinking subject”, i.e. the logical functions (Prol, AA 4:298). In other words, a judgement of perception is an act of thought that requires a certain use of the understanding, much as do the Lockean examples of the workmanship of the understanding. By contrast, judgements of experience are claims of the form: “x is the case”, e.g. “The air is elastic” (Prol, AA 4:300) and “If the sun shines on the stone, it becomes warm” (Prol, AA 4:301). Accordingly, they claim that something is the case (recall Kant’s appeal to the copula in §19 of the B-Deduction), not merely that it now seems to me to be the case. As such, judgements of experience lay claim to a two-fold projectibility: to other times and to other cognisers, which is to say that they are objectively valid. 

In line with this analysis, Kant characterises the objective validity of a judgement of experience as its “necessary universal validity” (Prol, AA 4:298), and he claims that “[o]bjective validity and necessary universal validity (for everyone) are interchangeable concepts” (Prol, AA 4:299). Necessity and universality are for Kant the twin criteria of the a priori. Moreover, they are likewise interchangeable, since whatever is necessarily true holds with strict universality and vice versa. But, once again, since Kant is here referring to a species of empirical judgement, it is clear that this necessity cannot be taken to mean that the judgement is necessarily true, which is why I have characterised it as normative. In fact, Kant underscores the normative nature of this necessity by pointing out that “there would be no reason why other judgments necessarily would have to agree with mine, if there were not the unity of the object—an object to which they all refer, with which they all agree, and, for that reason, must also harmonize among themselves” (Prol, AA 4:298.22–6). In other words, a merely de facto agreement with the judgement of others would not suffice to ground the objective validity of a judgement, since the claim is that the judgement of others ought to agree with mine; though, since the claim is normative, it does not imply that they actually will agree. I have claimed that this brings with it a kind of epistemic right to expect their agreement; though, again, this ‘right’ is highly defeasible, since my judgement, while objectively valid, might not be true. And for the same reason universality cannot here mean holding for every conceivable x. Moreover, Kant makes this clear when he characterises this universality as holding for everyone, that is, the universe of cognising subjects, thereby making it a subjective universality, rather than an objective universality that holds for the universe of objects cognised under a certain description, as would be the case with synthetic a priori judgements.

Since the decisive difference between judgements of perception and judgements of experience is that the former contain only empirical concepts, while the latter contain in addition to these a pure concept of the understanding or category, our next question is how this addition is supposed to account for the objective validity of the latter. Kant’s answer lies in the nature of the categories as concepts of an object in general. Otherwise expressed, they are the rules through which empirical concepts and the intuitions subsumed under them are related to an object. And since all concepts for Kant serve as rules, the categories are second-order rules or, rules for the legitimate use of the first-order rules that are provided by empirical concept. But, assuming that this is the gist of Kant’s answer, Allais asks why the empiricist should find any of this compelling. As she puts it, presumably speaking for the empiricist:

The fact that applying empirical concepts to objects presupposes that what they are applied to is already thought of as an object does not show that the latter requires a priori concepts, and that we could not, rather, have a conception of objects consisting entirely of empirical concepts that enables us to think of things as objects in the way needed for empirical concept application.

Inasmuch as the view that Allais here describes looks very much like an empiricistic version of conceptualism (if there is such a thing), I assume that she is speaking as a devil’s advocate, rather than proposing the non-conceptualist view that she endorses. But setting that aside, the basic problem with this view is that it helps itself to the concept of an object, even though it is not an empirical concept. This is clearly true for Kant, since for him the concept of an object is not the concept of something that is given, as it must be for an empiricist (think of the Sellarsian ‘myth of the given’), but a problem to be solved. And what makes this a problem is that in order to represent something as an object of a certain kind, which is done by bringing it under an empirical concept, it is necessary to recognise it as an object tout court, that is, as something distinct from one’s representations to which they refer. Thus, Kant asks rhetorically, “what is meant by the expression ‘an object of representations”, and he answers that, “this object must be thought of only as something in general=X, since outside of our cognition we have nothing that could be set over against this cognition as corresponding to it” (A104). The empiricist whose views comes closest to recognising this is Locke; since his concept of substance as the posited, but unperceived, substratum of the perceived primary and secondary qualities, may be considered a functional analogue of Kant’s conception of an object distinct from and corresponding to our representations. But the problem, as was noted above, is that Locke tried but failed to account for the empirical nature of this concept and refused to recognise any ideas/concepts that were not derived from experience.  

Allais also suggests that empiricists might be able to provide a conception of objects that is composed entirely of empirical concepts, but which would nevertheless enable us to think of things as objects to which empirical concepts could be applied. I find this suggestion problematic for two closely related reasons. First, as noted above, it helps itself to the non-empirical concept of an object to which concepts can be applied, that is, the concept of a subject of predication, which has an essential logical function but no empirical counterpart. Second, a concern with the conditions of the application of concepts (empirical or otherwise) reflects the discursive model of cognition according to which cognition consists in the application of concepts stemming from the understanding to sensibly given data; but while it is clear that Kant proceeded on the basis of this model, the empiricists, at least the empiricists whose views Kant was criticising in the Deduction, did not. Instead, they assumed what I have called a perceptual model, in which cognition consists in the immediate apprehension of a particular content that is before the mind, that is, a kind of seeing with “the mind’s eye”.[2] The canonical formulation of this model is Locke’s definition of knowledge as “the perception of the connexion and agreement, or disagreement and repugnancy of any of our Ideas” (Locke 1975:525). And since for the empiricists who analyse cognition on the basis of this model, perceptions are considered the objects as well as the vehicle of cognition, there is little conceptual space for the concept of an object of these perceptions to which concepts (if there are any) could be applied.  

(3) Since according to Kant “[e]very necessity has a transcendental condition as its ground” (A106), and “our thought of the relation of all cognition to its object carries something of necessity with it” (A104), it was necessary for Kant to provide the transcendental condition that grounds this relation, in the sense of warranting the claim of objective validity that is built into it. The condition that Kant supplies is transcendental apperception. By apperception in general Kant understands the power or capacity for self-consciousness, and by transcendental apperception a consciousness of oneself qua thinker, which he sometimes contrasts with empirical apperception, understood as an empirical consciousness of one’s mental state through inner sense. For Kant transcendental apperception is a transcendental condition of the possibility of experience, and, therefore, of forming and applying empirical concepts, because it is a condition of the possibility of thinking. And it is the latter because thinking for Kant is an inherently reflexive, self-conscious activity. Thus, while one might perceive or apprehend an x that is F, e.g. a stone that is warm (non-rational animals can do this) without any use of the understanding, to take an x as an F, which in the A-Deduction Kant calls “the synthesis of recognition in the concept” (A103), is not only something that one does, it also requires an awareness of what one is doing, an awareness that Kant refers to in both versions of the Deduction as a “consciousness of synthesis”.

Kant’s argument linking transcendental apperception, so conceived, to the categories and through them to experience constitutes the heart of the Deduction in both editions; and after devoting over 450 pages to the topic in my book I cannot pretend to do justice to it here. But at the risk of grossly oversimplifying an extremely complex line of argument, with regard to which Kant himself was never completely satisfied, I maintain that it turns ultimately on the assertion of a reciprocity between two unities: the unity of consciousness and the consciousness of unity. By the former is understood the transcendental unity of apperception, that is, the unity or, better, identity, of the thinking subject qua engaged in the act of thinking. The basic idea is that only a unified or single consciousness can represent to itself a multiplicity of representations as constituting a unity. In my earlier and more whimsical days I expressed this thought with the proverbial “It takes one to know one”. Moreover, since, as was noted above, this representing, i.e. thinking, is an inherently reflexive activity it follows that it is also necessary to be conscious of one’s identity qua thinker. In other words, one cannot unify a and b in a single consciousness without also being conscious of the identity of the I that thinks a and the I that thinks b. By the latter is understood the consciousness of an objective unity, that is, one that is the product of a unification or synthesis of representations by means of the categories, which are by definition concepts of an object in general or as such. Given this reciprocity, the argument proceeds by trying to show that the unification or synthesis of a cogniser’s representations through the categories is a necessary condition of the consciousness of self-qua cogniser, from which it follows (by the reciprocity thesis) that it is also a sufficient condition of the consciousness of an objective unity, that is, of an object corresponding to and distinct from the cogniser’s representations. Kant encapsulates this line of argument in the formula: “The a priori conditions of a possible experience in general are at the same time conditions of the possibility of objects of experience” (A111).

By the a priori conditions of a possible experience are here understood conditions of experiencing, which Kant elsewhere characterises as “subjective conditions of thinking”; and the question is whether these conditions also have objective validity, which Kant glosses as “yield conditions of the possibility of all cognition of objects” (A89/B122). By thinking is here understood discursive thinking, in which representations are referred to an object in an act of judgement. And by the subjective conditions of thinking are meant conditions of the possibility of unifying one’s representations in a single consciousness in such a way that the cognising subject can become consciousness of itself as the owner of this thought. While to claim that these conditions, i.e. the categories, have objective validity is to assert that they have an objectivating function, as the rules through which a relation to an object is thought by the human understanding. Moreover, since by the reciprocity thesis the conditions of the possibility of the representation of an object are at the same time conditions of the possibility of attributing these representations to a single thinker, it follows that representations that cannot be related to an object, cognitive outliers, as it were, cannot be brought to the unity of consciousness or apperceived. This does not mean that such representations are impossible; Kant was too attentive a student of Leibniz and his conception of small perceptions to assert that view. Rather, the claim is that they are “nothing for us”, cognitively speaking, “intuition without thought”, as he says on A111, which as such cannot be objects of a possible experience. To be sure, Kant does not express the point in this way, since what he typically denies being experienceable is the merely intelligible or noumenal on the grounds that it cannot be given in intuition. But inasmuch as experience for Kant requires concepts as well as intuition, what cannot be brought under a concept cannot be an object of possible experience.

Assuming that this, or something like it, is the basic argument of the Deduction, Allais’ main worry appears to be that it might not be compelling to the empiricist, whom she considers to be its main target. I agree that empiricism, that is, the classical empiricism of Locke, Berkeley and Hume, as well as their German counterparts, e.g. J.G. Feder, was Kant’s main target in the Deduction; but I hasten to add that it was not his only one. For though it has garnered much less attention, one of Kant’s concerns in the Deduction was to restrict the use of the categories to objects of possible experience. In fact, Locke was specifically criticised for not limiting them in this way, despite his claim that they have an empirical origin (see B127). Setting that aside, however, I believe that the focal point of the dispute between Kant and these empiricists is the concept of an object or, more precisely, the concept of an object in general or as such (überhaupt). While it is clear that these empiricists would concede that the application of concepts, or any act of thought for that matter, requires an (intentional) object, it seems that, with the partial exception of Locke, with his halfhearted appeal to the concept of substance as the unperceived substratum of the perceived qualities of things, they would reject the suggestion that it requires a distinct concept of an object (not to mention the concept of an object in general or as such). Rather, for these empiricists cognition begins with something that is already considered as a determinate object, e.g. an impression of redness or, in the Sellarsian mode, which seems appropriate here, a red impression, or an impression of hardness, a.k.a a hard impression.

As already noted, however, Kant’s view is quite different. Indeed, since the categories, which are the a priori concepts at issue in the Deduction, are introduced by Kant as concepts of an object in general or as such, it could be argued that virtually the whole weight of its anti-concept empiricism argument falls ultimately on it. Moreover, Kant makes this clear when, in the introductory portion of the Deduction that is contained in both editions, he writes: 

The question now is whether a priori concepts do not also precede, as conditions under which alone something can be, if not intuited, nevertheless thought as an object in general, for then all empirical cognition of objects is necessarily in accord with such concepts, since without their presupposition nothing is possible as object of experience. (A93/B125–6)

This passage immediately follows Kant’s reminder that appearances necessarily conform to the formal conditions of sensibility, i.e. space and time, because the Aesthetic has shown that it is only through these forms that appearances can be given in sensible intuition. But rather than inferring directly from this that appearances must likewise conform to the conditions of thinking, which would make the Deduction too easy, he presents this conformity as precisely the question that the Deduction must resolve. Interestingly, this problem, at least as Kant presents it, does not arise for the empiricists, because, apart from the core requirement that all the materials for thinking be supplied by experience and the minimal constraints imposed by formal logic, the idea of conditions of thinking would have been completely alien to them. It does unavoidably arise, however, if, with Kant, one considers cognition as a discursive activity; for then the blindness of intuitions without concepts means that what is immediately apprehended through the senses is conceptually indeterminate (a spatio-temporal = x), which Kant makes clear with his characterisation of appearance as “[t]he undetermined object of an empirical intuition” (A20/B34). And this, in turn, means that the understanding is at work at the ground floor of cognition, so to speak, taking what is perceived as red or hard, rather than merely having a red or hard impression. Moreover, in order to take something apprehended as red or hard it is also necessary to take it as a something, that is, as an object, which involves an appeal to the concept of an object. In other words, while objects may be given in sensible intuition for Kant, they are not given as objects. Rather, they are taken as objects, which, as an act of the understanding, requires that they be brought under the concept of an object in general. And since the categories for Kant are concepts of an object in general, the move from this concept to the categories is short; though it is, of course dependent on the prior assignment of this status to the categories, which is the task of the Metaphysical Deduction.

I would also like to respond to a more specific criticism raised by Allais that bears on the issue of Kant’s reply to the empiricists. It concerns my reading of Kant’s claim that “[e]very intuition contains a manifold in itself, which however would not be represented as such if the mind did not distinguish the time in the succession of impressions on one another; for as contained in one moment no representation can be anything other than absolute unity” (A99). I took Kant to be there denying the temporal analogue of Berkeley’s conception of a minimum visibile on the grounds that a moment (“ein Augenblick”) is a limit rather than a part of time. Allais replied that this is not needed, since there are other ways to make sense of this passage. Although this may be true, I believe that Kant’s insistence on the infinite divisibility of time (and space) gave him a sufficient reason to deny it. And, non-coincidentally, this is also a conception with which Berkeley and Hume had considerable difficulty dealing with successfully, since conceptions like infinite divisibility do not seem to fit their perceptual model of cognition.  

Finally, it seems that at the heart of Allais’ criticisms or, perhaps better, her reservations regarding the limitations of my account of Kant’s arguments, is that I have a propensity to preach to the choir, that is, to address my analyses to those who are already sympathetic to Kant, though perhaps confused or uncertain on points of detail; and that as a result I have relatively little to say that the hardcore empiricist might find persuasive. And I suspect that there may be some truth to this; particularly since my main goal in writing about Kant has always been to try to make sense of it for myself. Nevertheless, I also think, and believe that perhaps Allais may concur, that, given the widespread disagreement among members of this choir, that is, would-be Kantians, such preaching may be of some use. 


[1] In the book (p. 303) I argue that in the Prolegomena Kant uses the conception of a judgement of perception didactically to indicate how an empirical judgement without the categories would have to be understood.

[2] I analyse the perceptual model with particular reference to Hume in Allison (2008).




Allais, L. (2015), Manifest Reality, Kant’s Idealism & His Realism (Oxford: Oxford University Press).

Allison, H. (2008), Custom and Reason in Hume, A Kantian Reading of the First Book of the Treatise (Oxford: Clarendon Press).

Ameriks, K. (1978), ‘Kant’s Transcendental Deduction as a Regressive Argument’, Kant-Studien 69: 273–87.

Hume, D. (1999), An Enquiry concerning Human Understanding (Oxford: Oxford University Press).

Locke, J. (1975),  An Essay concerning Human Understanding (Oxford: Clarendon Press).

© Henry E. Allison, 2019.

Henry Allison is Emeritus Professor of Philosophy at the University of San Diego and Boston University. He has published widely on Kant, Spinoza and German Idealism and is best known for the standard work of Kant scholarship Kant’s Transcendental Idealism. An Interpretation and Defense (Yale UP 1983, 2004). His next major work will be published by Cambridge University Press in late 2019, under the title of Kant’s Conception of Freedom. A Developmental and Critical Analysis.