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


By Alison Laywine

The new book by Henry Allison, under discussion here, is the result of an engagement over more than forty years with Kant’s Critique of Pure Reason. It is Allison’s first book-length treatment of the Transcendental Deduction: most welcome, indeed, because it is an opportunity for him to elaborate his understanding of the Deduction beyond the smaller compass of Part Two of Kant’s Transcendental Idealism: an Interpretation and Defense (2004). He carefully examines not just the presentation of the Deduction in the second edition of the Critique from 1787, but also that of the first edition from 1781. He supplements his examination with a careful review of Kant’s writings relevant to the Deduction in the six years between the two editions: not just the Prolegomena, but also Reflexionen 5923–35 and the famous note about the Deduction in the preface to the Metaphysische Anfangsgründe der Naturwissenschaft.

Moreover, the first three chapters of his book attempt to review and assess the prehistory of the Deduction, as documented by Kant’s various writings from the early 1760s: those that testify to what Allison calls his transition “from straightforward metaphysics in the traditional sense […] to a methodological concern with the nature and conditions of the possibility of metaphysics” (p. 3). The historical review continues to 1770 with the Inaugural Dissertation and then into the ‘silent decade’ with an examination of the loose notes from the time that seem to record Kant’s efforts to overcome difficulties in the Dissertation and move beyond it.

Allison thus characterises his book in its subtitle as an ‘analytical-historical commentary’. This could mean different things; perhaps we could ask Allison to step back from all the details and reflect on what he takes such a commentary to be. An ‘interpretation and defence’ is self-explanatory. My impression, as his reader, is that an ‘analytical-historical commentary’ is a study that proceeds step-by-step. The path is laid out in three different modes at once: chronologically, textually and argumentatively. Each step takes us closer in time to the appearance of the second edition of the Critique in 1787; each step also sets out a claim or a paragraph closer to the end of a significant stretch of text and/or a premise closer to the conclusion of an argument. There is an advantage to proceeding in this way: it invites the reader to pause at each of Kant’s stations on the road to the last sentence of §27 of the B-Deduction. It is a stimulating and illuminating journey. But it has at least one disadvantage.

The more straight and narrow the path to the terminus, the harder it may be to step back and put elements of the larger lessons in perspective: we may see everything there is to see of a passing landscape, from the window of our train, as we travel from one city to another, without fully understanding the historical or geographical significance of the region through which we are making our way. I do not mean to imply that Allison is insensitive to the larger lessons raised by his commentary. On the contrary, he calls attention early to developments that he takes to point ahead to matters of philosophical significance. For example, in Chapter 3, he takes Kant to be trying, in the so-called Duisburg Nachlaß from around 1775, to be working out a way of conceiving the relation between sensibility and the understanding that points forward to the Schematism. Still, because of the doggedly stepwise presentation of the book, it can be hard for the reader to see in detail how certain large issues should be understood. In the interest of stimulating discussion, I would like to focus on just one—and, then again, incompletely. I trust that Allison will appreciate an opportunity to step back from his commentary and fill us in on it, from the bird’s-eye perspective.

The main issue I shall signal for discussion is that of ‘normativity’. It is an issue that percolates through much of the book—the latter half of it, anyway. Allison appeals to the notion to clarify the so-called ‘argument from below’ in the A-Deduction (Chapter 6); he takes it to be driving the sections relevant to the Deduction in the Prolegomena (Chapter 7); he detects it at work again in the B-Deduction (Chapters 8 and 9). If Allison is right, it is a notion central to Kant’s thinking about experience and the categories. But what does it add up to? Perhaps the way to approach this question is to consider first Allison’s motivations for suggesting it (on Kant’s behalf).

It is supposed to resolve, perhaps among other things, an apparent paradox in remarks that Kant makes about experience and empirical judgements. On the one hand, he says in many places that empirical cognition lacks necessity. For example, we find the following pronouncement in the introduction to the second edition of the Critique: “Experience does indeed teach us that something is thus or such, but not that it could not be otherwise” (B3). The passage continues:

Experience never gives its judgements true or strict universality, but only the presumptive [angenommene] and comparative kind (through induction) so that we may properly say only this: however much we have perceived up to now presents no exception to this rule or to that. If, then, a judgement is thought in strict universality, i.e. so that no exception be allowed possible, it is not derived from experience, but is valid absolutely a priori. Empirical universality is, then, but an arbitrary heightening of the validity from that which holds for the most part to that which holds always and in every case […]. (B3, all translations are my own)

The passage just quoted focuses on universality. But it is relevant for present purposes because, if empirical cognition lacks strict universality and thus admits exceptions to the relevant rules, it will lack strict necessity as well. For example, it will not necessarily be the case that precipitation always occurs when barometric pressure falls, because exceptions have occasionally been recorded. Now the problem is that our passages from the introduction to the second edition of the Critique appear to be at odds with other passages that nevertheless associate necessity with all our knowledge—even that which is empirical.

In the A-Deduction, for example, Kant says that

[w]e find, however, that our thought of the relation of all knowledge to its object carries with it a certain necessity, this object is considered to be that which stands over against it so that our knowledge may not be arbitrary or anyway-you-please [aufs Geratewohl oder beliebig] but rather determined in a certain way a priori (A104).

The claim here is supposed to apply to all knowledge. If we accept the claim and accept, as we must, that some of our knowledge is empirical, it will follow that some kind of necessity attaches to some of our empirical thought and judgements. A passage at least superficially like this one can be found in §19 of the B-Deduction; it appeals to some kind of necessity to distinguish between the ‘subjective unity of apperception’ underlying associations carried out by the imagination and the ‘objective unity of apperception’ that somehow makes all judgement possible. Some judgements are empirical. But if they too depend on the ‘objective unity of apperception’, they will exhibit the relevant necessity. This confronts us once again with a claim seemingly at odds with the remarks we considered earlier from the introduction to the second edition of the Critique.

We know that Kant was aware of the tension, because he sought to dispel it. The evidence can be found in a footnote to §22 of the Prolegomena where he writes:

But how does the claim that judgements of experience are supposed to contain necessity in the synthesis of perceptions square with my claim, frequently emphasised above, that experience, as knowledge a posteriori, can only yield contingent judgements? (Prol, AA 4:305.23–6)

Kant tries, in the continuation of the footnote, to sort things out for us. But what he says there is not, in my view, helpful. Nor is it useful for the task at hand. The task at hand is first to indicate that, if Allison is right, Kant’s effort to sooth the tension depends on the relevant understanding of normativity, and second to ask him (Allison) to clarify how he himself understands this notion. In the interest of carrying out the first task first, I fear that I must clumsily anticipate questions associated with the second task. Let me start by briefly collecting some of the suggestions about normativity that Allison makes in the book.

The first thing to note is that Kant sometimes characterises concepts as rules—not, so far as I am aware, in the B-Deduction, but certainly in the A-Deduction, as when he says, for example, that “all knowledge requires a concept […], but [every concept] is always something universal, with respect to its form, and that serves as a rule” (A106). Rules can be more or less strict, but if they are rules at all, they require or demand something: if you wish to win a game of chess, then you must move the pieces in the prescribed way and you should try to command the centre of the board early on. As such, rules express a certain necessity. We may ask, though, what sort of necessity Kant has in mind? Allison denies that it is logical: I think he means that it is not (or not simply) the requirement that we avoid contradictions and draw valid inferences. He also denies that it is causal. I think this is supposed to ensure clarity of exposition: causal connections are indeed necessary, on Kant’s view, but though every concept is a rule and thus expresses some kind of necessity, not every concept is causal—hence, the necessity expressed by every concept as a rule must be distinct from the causal necessity expressed by causal concepts. At least this is what I take to be the upshot of Allison’s discussion on pp. 191, 222 and 302. If I have misunderstood, he should correct me. In any case, the upshot is supposed to be that the necessity at issue is ‘normative’. Allison writes as follows:

The essential point is that as rules concepts are themselves normative in the sense that they involve a kind of epistemic right to expect that whatever objects fall under the rule contain the properties that are expressed in the concept-rule and to require that other cognizers, who possess the concept will agree with one’s judgement. (p. 237)

This remark is not elaborated, though it is repeated elsewhere in the book. For we read in the discussion in Chapter 8 devoted to §19 in the B-Deduction that “[normative necessity] concerns how one ought to think and […] this is intimately linked with Kant’s conception of concepts as rules, which have a built-in normative force” (p. 365). Allison invites us to consider a Kantian example, namely the judgement that bodies are heavy (remember that a judgement, for Kant, is a concept put to work):

Here the concept of body functions as a first-order rule requiring one to predicate weight of anything that is brought under this concept. The necessity is normative in the sense that it imposes a requirement on anyone who applies the concept in a judgement. (p. 365)

I am not confident that I understand the nature or strength of the requirement imposed on me and my fellow ‘cognisers’ by the concepts we use. I could surely have a certain concept; objects could surely ‘fall under it’, and yet it could happen that these objects do not have all or any of the properties my concept predicates of them. I might have a certain concept that enables me to recognise an object as a certain natural kind, but my concept could be mistaken in various ways about what makes that kind the kind that it is. For example, I possess the concept of giant panda. I know that giant pandas fall under my concept, and I even succeed, with the help of my concept, in dependably distinguishing giant pandas from lesser pandas. But it was always part of my concept that giant pandas are a kind of raccoon, i.e. a member of the Procyonidae. It turns out that I am mistaken, for current science tells us that giant pandas are bears (Ursidae). So I do not have the right to expect that giant pandas have all of the properties my concept predicates of them, and I certainly have no right to expect you to be a party to my error. The non-logical, non-causal requirement that seems appropriate in light of this case is that we all be willing to correct our concepts and judgements in light of the best evidence at hand.

No doubt, Allison does not wish to be taken this way. I do not mean to be uncharitable, but only to take this opportunity to invite him to clarify. Until he takes up the invitation, perhaps the thing to say on his behalf and Kant’s is something like this.

It starts with the Kantian idea that our understanding is a spontaneous faculty. How you use the spontaneity of your understanding is entirely up to you. But there are certain constraints, not all of which may be characterised as ‘normative’—at least not if we invoke the notion of normativity only in those cases where we stand before choices about how to conduct ourselves, intellectually or otherwise. Our inescapable finitude, for example, places a very significant constraint: it compels us to take up a manifold given to us in sensibility. No finite understanding can produce a manifold of its own. Here there is no room for discretion, hence no cause to speak of normativity. We also have no choice but to make use of concepts. The understanding is a faculty for thinking; the concept is its currency. But how these concepts are formed and even how they are applied will be matters of discretion. They will nevertheless be subject to constraints. Here normativity may well kick in, if the constraints guide our discretion without eliminating all room for it. This is admittedly vague. But it may well be enough to dispel the seeming tension I indicated earlier between Kant’s claim in the introduction to the second edition of the Critique that anything derived from experience is contingent and his remarks in both versions of the Deduction that all knowledge exhibits a certain necessity, even that which is empirical. There is no tension here, Allison will argue—if I have properly represented his case in outline—because the necessity exhibited by empirical knowledge is soft, enough so to accommodate the contingency of whatever we derive from experience.

I take myself at this point to have completed as much of the first task I set myself as I require for present purposes. Along the way, I have already had occasion to take up elements of the second, which is to consider what Allison, speaking for Kant, takes normative necessity to be. I conclude by pursuing a line of questions about normative necessity that flows out of something I said earlier on that score, namely that normative necessity is soft. Now I would like to ask: just how soft is it?

Even in the game of chess, which I earlier took as a representative showcase for normativity, some of the rules are hard and non-negotiable. For example, you may not castle if your king is in check. But some of them are soft, by comparison. For example, it is not a good idea, in the earliest stages of a game, to move a piece more than once. This is a soft rule, because you can violate it, without ceasing to play chess altogether. The Alekhine defence violates it, because 1.e4, Nf6 exposes the black knight to capture if (predictably) White next advances its pawn to e5; to avoid capture, the black knight will have to move again. This defence is not a departure from chess; on the contrary, it can be justified within the bounds of the game. Its point is to tempt White to overextend its pawns. Black may even win the game, as Alekhine himself beat Steiner in Budapest in 1921. Chess is not the only field in which we may detect degrees of normative necessity. There are hard rules and soft rules in Renaissance polyphonic counterpoint. For that matter, the same may well be true in Kantian ethics. The requirement to acknowledge the moral law is a hard rule. You cannot deny that the moral law has a claim on you without ceasing to play the game of morality altogether. The duties that flow from the moral law will be softer than the requirement that we acknowledge it. Kant treats the duty not to lie as a perfect duty: you may not lie, and that is all there is to say about it. But if you did tell a lie, you would not, on that account, have ceased playing the game of morality. You would have played a very bad move that will cost you universal disapprobation among your fellow players. The duty to develop your talents may perhaps be softer than the duty not to tell a lie. You will incur disapprobation if you never develop any of your talents. But which one, to what degree and at what stage of your life? All of that is up for grabs. So let’s just call that one softer still.

If thinking is a field of normativity for Kant, then surely more needs to be said about normative necessity. However soft it may be by comparison with other kinds of necessity, it surely comes in degrees. How hard or soft are its requirements? The answer to that question will naturally depend on what we take those requirements to be and, just as important, how we think they relate to one another. I am willing to take seriously, at least as a claim about Kant’s position on these matters, that logic lays down a lot of these requirements and that it is normative in the relevant sense. Kant says so explicitly in §1 of the Jäsche Logic:

In logic, we wish to know not what the understanding is like and how it thinks and how it has proceeded up to now in thought, but how it ought to proceed therein. (Log, AA 9:14)

The idea that logic is normative for Kant has been developed by others, notably John MacFarlane (2002). It is controversial: it has been called into question by Clinton Tolley (2006). But I wish to take it on board for the purposes of this discussion, just because I find it highly plausible. Its interest for us is its implication that normative necessity includes logic and that we will not really understand what kind of necessity this is for Kant by trying to carve a space for it, as Allison does, between logic and causality. My quicky (programmatic) proposal is this: Kant’s considered view is that the requirements of logic are soft requirements that somehow presuppose, or come into play, once the hardest requirements on thinking are in place. The hardest requirements on thinking will be like the hardest rules of chess: you cannot violate these rules without ceasing to think. But so long as you abide by these rules, you will need the guidance of soft rules—those of logic—to regulate the use of your intellectual spontaneity so that you can make moves in the game. What are the hardest rules of thinking? Well, surely the principle of the synthetic unity of pure apperception, which we learn is

the highest point to which we must affix all use of the understanding, even the whole of logic, and subsequent to it, transcendental philosophy: indeed, this capacity [Vermögen] is the understanding itself (B134n.).

My proposal implies that a fuller account of thinking as a normative activity would need to work out more precisely how logic and the principle just mentioned relate to each other, i.e. it would have to answer the question how the (comparatively) hard gets softened around the edges.

There is, of course, much more to say about normative necessity. My reflections have taken us only as far as the principle of the synthetic unity of apperception and its relationship with logic. Then again, all I have done, even on that score, is sketch a few programmatic preliminaries. Normative necessity is an important issue in Allison’s book. But the book is about the whole Transcendental Deduction, the revisions it underwent and its prehistory in the 1760s. The reader will discover that many other issues come into play. I have had the luxury of focusing on just this one, because we have two other contributors, and they will no doubt take up other issues and questions. In the meantime, I shall now turn things over to Allison and invite him to tell us more about how understands the issue of normative necessity.


Allison, H. (2004), Kant’s Transcendental Idealism: an Interpretation and Defense, revised & enlarged edition (New Haven: Yale University Press).

MacFarlane, J. (2002), ‘Frege, Kant, and the Logic in Logicism’, Philosophical Review 111 (1): 25–65.

Tolley, C. (2006), ‘Kant on the Nature of Logical Laws’, Philosophical Topics 34 (1/2): 371–407.

© Alison Laywine, 2019.

Alison Laywine is Associate Professor of Philosophy at McGill University, Canada. She is the author of Kant’s Early Metaphysics and the Origins of the Critical Philosophy (Ridgeview 1995).