By David Landy
Before attempting to answer the insightful and challenging questions that Gomes and Jankowiak have raised, I would like to begin by thanking them both for engaging so deeply and productively with my work. Owing as it does so much to Kant and Sellars, and limited as it is by my own expository and explanatory abilities, Kant’s Inferentialism is an admittedly difficult book. Thus, I appreciate all the more Gomes’s and Jankowiak’s successful efforts to plum its depths and emerge with questions and concerns that get at the very core of its project. In what follows, I attempt to answer and address those.
1. Gomes on Kant’s Genealogical Idiom
Gomes raises the question of the role of Kant’s adoption of the language of genealogy with respect to the concepts of necessary connection and the external world, and rightly notices that on an inferentialist understanding such as mine, this language appears out of place. Whereas it makes sense to ask after the genealogy of a thing, e.g. an idea in Hume’s sense, it makes less sense to ask after the genealogy of a rule. One can expect to find the causal origin of an idea, e.g. that it is derived from an impression, because an idea is a naturalistic object governed by causal laws. A rule of inference is not a naturalistic object (though it must be embodied by certain regularities of performance of naturalistic objects, empirical selves), and so seeking its causal origin appears to be a misguided endeavour. Thus, anyone who understands Kant’s theory of concepts as containing an essentially normative element appears to be in an awkward position in explaining his talk of genealogy. Fair enough.
Broadly speaking, I think that the correct approach to resolving this awkwardness has two related parts. First, I would suggest that Kant’s use of the idiom of genealogy serves mostly a negative function. It is primarily his way of eschewing the genealogical approach of his predecessors. His most fundamental point when he uses this language is that our a priori representations do not derive from experience. That they do derive from the understanding is a useful, but misleading, metaphor. Secondly, it is worth noting that Kant himself often shifts his discussions concerning the genealogy of our a priori representations, which Gomes rightly recognises is concerned first and foremost with the Possession Problem, to questions of the justification of the use of such representations. He does so because as he understands them genealogical questions are often in the service of justificatory ones—recall that in a legal proceeding the quid juris is often answered by tracing a claimant’s right to some property through its history of ownership to its original acquisition—but in the case of the pure a priori concepts, this form of justification is misplaced. We do not answer the quid juris regarding a priori concepts by tracing their origin at all, or so I have argued. The justification of the pure a priori concepts is not backward looking, so to speak, but forward looking. We justify the use of those concepts not by appeal to their source, but rather by appeal to what they allow one to do (represent oneself as the single subject of experience persisting through time). So while Kant often begins writing in a genealogical idiom, he just as often transitions away from investigating the origin of our concepts by making the underlying justificatory question explicit, and then answering that questions in a way that makes little or no use of genealogy at all.
To bring out these two aspects of Kant’s use of genealogical language—that he uses it to make a negative point against those who take the origin of all representations to lie in experience, and that he is more concerned with the justificatory question motivating the interest in the genealogy of our representations—we need look no further than the very passages that Gomes himself cites. Consider the passage that Gomes quotes from Kant’s response to Eberhard:
The Critique admits absolutely no implanted or innate representations. One and all, whether they belong to intuition or to concepts of the understanding, it considers them as acquired. But there is also an original acquisition (as the teachers of natural right call it), and this of that which did not yet exist at all, and so did not belong to anything prior to this act. (ÜE, AA 8:221)
The first two sentences here are addressed, in some sense, to a genealogical question: whether our representations have their origin in God or in us, or are acquired. But in the third sentence, Kant engages in a bit of wordplay to change the subject. Whereas a moment before he uses ‘acquired’ (erworben) to indicate that our a priori representations are not innate, he next uses it as part of the phrase ‘original acquisition’ (ursprüngliche Erwerbung), which as he indicates, is a term from legal writing. Like ‘deduction’, ‘original acquisition’ comes from Deduktionsschriften, and refers to the acquisition of a property that has never been the property of anyone else, i.e. the creation of an entirely new proprietary right. This new right, in turn, can arise in two different ways. It can be a right to something that did not previously exist, e.g. the copyright to which an author is entitled over his or her work, or it can be a right to something that had not previously been owned, e.g. a gold prospector’s staking a claim to unoccupied land. In this case, Kant is explicit that the representations with which he is concerned are of the former kind: their original acquisition is our right to “that which did not yet exist at all, and so did not belong to anything prior to this act”. If, however, our right to employ a priori representations is an original acquisition of something that did not exist prior to our own spontaneous activity, this is as much to say that there is no genealogy available for these representations, and the question of the legitimacy of their use must proceed in another direction. I argue that it proceeds by showing that a priori representations are the conditions of representing the world as consisting of the necessary connections among alterations in substance, which is in turn necessary for representing oneself as the single subject of experience persisting through time. Kant goes on:
[F]or neither of these does our cognitive faculty get from objects as given therein in-themselves, rather it brings them about, a priori, out of itself. There must indeed be a ground for it in the subject, however, which makes it possible that these representations can arise in this and no other manner, and be related to objects which are not yet given, and this ground at least is innate. (ÜE, AA 8:221)
Again we see Kant’s two-part strategy. First, what is important about the genealogical question is its negative answer: “neither of these does our cognitive faculty get from objects as given therein in-themselves.” Second, Kant pivots from the genealogical question to its underlying justificatory motivation: “There must indeed be a ground” that makes it possible that these representations “be related to objects which are not yet given”. The ground of the relation of a priori representations to objects from which they are not yet derived is precisely the place that Kant locates the quid juris as opposed to the quid facti of the Deduction. Thus we see both parts of the strategy that I outlined above at work here. Kant begins by addressing the genealogical question raised by his opponent, but quickly pivots by denying the relevance of the genealogical issue, and then switching to addressing what he takes to be the motivation underlying the genealogical question: the Justification Problem. It is precisely because that there is no justificatory answer to the genealogical question that the justification problem is so pressing and receives the unique kind of answer—a transcendental deduction—that it does.
Again, we can see the same moves at work in the other passages that Gomes cites. The first is from B159, where, as Gomes notes, Kant writes that “[i]n the metaphysical deduction the origin of the a priori categories in general was established”. Presumably it is Kant’s use of ‘origin’ here that captures Gomes’s eye, but his truncation of Kant’s complete sentence belies what I take to be Kant’s more complicated anti-genealogical approach:
In the metaphysical deduction the origin of the a priori categories in general was established through their complete coincidence with the universal logical functions of thinking, in the transcendental deduction, however, their possibility as a priori cognitions of objects of an intuition in general was exhibited (§§ 20, 21). (B159)
What is “the origin of the a priori categories in general” that Kant established in the metaphysical deduction? It is their “complete coincidence with the universal logical functions of thinking”. That is a strange genealogy indeed! Rather than purporting to trace the history of the a priori categories to their origin in the understanding, Kant cites the fact that they are identical to the logical forms of judgement. That is, Kant avoids the genealogical question altogether by citing the very fact that vitiates the propriety of that question. Reading Kant as instead offering an answer to a genealogical question makes what that answer would be seem utterly bizarre. It is as if you ask me who Leland’s parents are, and I answer, “Leland is Bob”. We can imagine a scenario in which that is an informative answer—e. g. where I know that you know who Bob’s parents are—but the answer itself is hardly a paradigm of genealogy. Rather, what I would suggest is that Kant’s answer to the genealogical question is a way of showing why such questions are misguided. So imagine instead that you ask me who Leland’s parents are because you want to know who raised him to be such a jerk. In response, I point out that Leland is an autonomous agent who is responsible for his own actions no matter his genealogy. That’s more like what Kant is after. What is the origin of the categories? No, no. It is not their origin that matters, because their use will not be justified by appealing to that origin. Rather, the categories are the meta-conceptual rules that describe the forms that any intuition must take in order to be a representation of a determinate object—i.e. the categories completely coincide with the logical functions of thinking—the representation of which is itself the means by which one represents oneself as a single subject of experience persisting through time.
This explanation of the ‘origin’ of the categories is helpful for understanding the last of Gomes’s citations, this time from A81/B107 where Gomes notes Kant’s reference to the table of categories as an “ancestral registry of the understanding”. Again, that appears to be an unmistakably genealogical idiom, but a closer examination of the text reveals Kant switching quickly from genealogy to the picture of the categories as meta-conceptual sortals. Here is the passage that Gomes cites, in which Kant laments Aristotle’s haphazard compilation of his list of categories.
Aristotle’s search for these fundamental concepts was an effort worthy of an acute man. But since he had no principle, he rounded them up as he stumbled on them, and first got up a list of ten of them, which he called categories (predicaments). Subsequently he believed that he had found five more of them, which he added under the name of post-predicaments. But his table still had holes. Further, it also included several modi of pure sensibility (quando, ubi, situs, as well as prius, simul,) as well as an empirical one (motus), which do not belong in this ancestral registry of the understanding; derivative concepts were also included among the primary ones (actio, passio), and several of the latter were entirely missing.
As Kant continues, it becomes clear that he is not thinking of this ancestral registry as a genealogy at all, but rather as a specification of the species and genera of various concepts. When Kant says in the continuation of that passage (below) that he is after, “the family tree of pure understanding”, this is clearly a play on a Porphyrian tree, rather than a demand for a genealogical account. Kant is, after all, discussing Aristotle’s categories here, and it is in his introduction to that work, Isagoge, that Porphyry first presents his eponymous tree as a way of understanding them. Kant writes further:
For the sake of the primary concepts it is therefore still necessary to remark that the categories, as the true ancestral concepts of pure understanding, also have their equally pure derivative concepts, which could by no means be passed over in a complete system of transcendental philosophy, but with the mere mention of which I can be satisfied in a merely critical essay.
Let me be allowed to call these pure but derivative concepts the predicables of pure understanding (in contrast to the predicaments). If one has the original and primitive concepts, the derivative and subalternate ones can easily be added, and the family tree of pure understanding fully illustrated. Since I am concerned here not with the system, I reserve this supplementation for another job. But one could readily reach this aim if one took the ontological textbooks in hand, and, e.g., under the category of causality, subordinated the predicables of force, action, and passion; under that of community, those of presence and resistance; under the predicaments of modality those of generation, corruption, alteration, and so on. The categories combined either with the modis of sensibility or with each other yield a great multitude of derivative a priori concepts, to take note of which and, as far as possible, completely catalogue would be a useful and not unpleasant but here dispensable effort.
To subordinate the predicables of force, action, and passion under that of causation, etc. is not to give an account of the origin of these concepts in that of causation, but is rather to present the former as a species of the genus that is the latter, and to specify their differentia. That is exactly what a Porphyrian tree accomplishes, it is not genealogical, and it is exactly what an interpretation of the categories as meta-conceptual sortals would predict. Thus it again turns out that the context of the citation that Gomes gives shows that Kant’s adoption of the idiom of genealogy represents nothing more than his adopting the façon de parler in order to reject the conception of justification that underlies it, and point instead to the proper methodology. What justifies the employment of the categories is that the categories are nothing other than the concepts of the forms that intuition must take in order to appear in objectively valid judgements, and we are justified in making objectively valid judgements because doing so is the means by which one represents oneself as a single subject of experience persisting through time. The “origin” or “ancestry” or “family tree” of the categories is not given by tracing any actual genealogy of them, but instead by showing their “complete coincidence” with the logical forms of judgement.
2. Gomes on Preformation Systems
Gomes’s second concern is that in casting the justification of the pure a priori concepts in this way, the inferentialist proposes what amounts to a preformation system of the understanding, i.e. that the justification of the employment of the categories to objects of experience is not grounded in those objects themselves, but rather only in certain necessary features of the experiencing subject. As Gomes puts it,
as meta-conceptual rules of inference, the categories structure what it is for something to be an object such that no sense can be attached to the question of whether objects really are as we necessarily represent them as being.
Because the categories are the meta-conceptual sorts that describe what form an intuition must take if it is to appear in a judgement with objective validity, any object represented by such an intuition, any object of possible experience, any appearance, must embody the categories. For example, since we can make universal, particular, or singular judgements, intuitions must be representations of unities, pluralities, or totalities. Gomes’s concern about that line is that
the inferentialist account is able to make this move only by grounding the correspondence between our employment of the categories and their applicability—between, that is, the forms of thinking and the laws of nature—in the kinds of inferences which we are so constituted as to make. And this looks like the kind of grounding base which Kant warns us against in §27 of the Deduction.
Before continuing I want to pause to issue a warning. What Gomes gets right about the inferentialist interpretation is that it grounds the applicability of the categories in certain necessary features of our representative constitution. That alone, however, cannot be sufficient for counting that interpretation as a kind of preformation system because Kant himself is explicit that the only possible grounding of the employment of the categories will take exactly this form:
There are only two possible cases in which synthetic representation and its object can come together, necessarily relate to each other, and, as it were, meet each other: Either if the object alone makes the representation possible, or if the representation alone makes the object possible. If it is the first, then this relation is only empirical, and the representation is never possible a priori. And this is the case with appearance in respect of that in it which belongs to sensation. But if it is the second then since representation in itself (for we are not here talking its causality by means of the will) does not produce its object as far as its existence is concerned, the representation is still determinant of the object a priori if it is possible through it alone to cognize something as an object. (A92/B124–5)
As I see it, the inferentialist interpretation is precisely a way of understanding Kant’s claim that “the representation is still determinant of the object a priori” insofar as “it is possible through it alone to cognize something as an object”. As meta-conceptual sortals, the categories specify the forms that an intuition must take if it is to be a representation of an object. It is in that sense that through them alone is it possible to cognise something as an object. Where representations do not take one of these forms, they are not cognitions of objects at all.
So, while the inferentialist does make an essential appeal to the necessary representative constitution of the experiencing subject, this alone is not sufficient to make inferentialism a kind of preformation system. In fact, if we are to believe Kant, making such an appeal is the only way to account for the possibility of synthetic a priori representations. What makes something a preformation system is that this appeal to the necessary representative constitution of the experiencing subject also severs the tie between those features of the subject and the features represented by that subject in the object. Here is Gomes’s summary of Kant’s warning:
On preformation views, it is genuinely necessary that we employ the categories, but the necessity of our doing so is ultimately grounded in aspects of our own subjective dispositions and not on the relation between the categories and the objects of experience. And if we cannot show more than this merely subjective necessity, then we cannot justify our use of the categories: Hume’s scepticism about justification is sustained.
The problem with a preformation view is that there is a disconnect between the ways that we necessarily represent the world, and the way the world is. Even if it is necessary for us to represent the world in the way that we do, and even if in so doing we thereby represent the world the way it is, because our doing so is not grounded in the way the world is, our representation cannot be justified in the proper way. The problem with a preformation system is that while it is necessary for the subject the represent the world in some way, and the subject might just so happen to thereby represent the world correctly, there is no necessary connection between the way the subject represents the world and the way the world is. It is exactly this missing element of necessity that Kant cites in his description of preformation systems:
Now there are only two ways in which a necessary agreement of experience with the concepts of its objects can be thought: either the experience makes these concepts possible or these concepts make the experience possible. […]
If someone still wanted to propose a middle way between the only two, already named ways, namely, that the categories were neither self-thought a priori first principles of our cognition nor drawn from experience, but were rather subjective predispositions for thinking, implanted in us along with our existence by our author in such a way that their use would agree exactly with the laws of nature along which experience runs (a kind of preformation-system of pure reason), then […] this would be decisive against the supposed middle way: that in such a case the categories would lack the necessity that is essential to their concept. For, e.g., the concept of cause, which asserts the necessity of a consequent under a presupposed condition, would be false if it rested only on a subjective necessity, arbitrarily implanted in us, of combining certain empirical representations according to such a rule of relation. I would not be able to say that the effect is combined with the cause in the object (i.e., necessarily), but only that I am so constituted that I cannot think of this representation otherwise than as so connected; which is precisely what the skeptic wishes most, for then all of our insight through the supposed objective validity of our judgments is nothing but sheer illusion. (B167–8; boldface mine)
Kant repeatedly stresses here that the problem with preformation systems is that there is no necessity between the way that the subject represents the object and the way that the object actually is. “In such a case the categories would lack the necessity that is essential to their concept.” Notice that the failure that Kant finds with the role of the concept cause in a preformation system is that it would rest only on a subjective necessity. It is not the subjective necessity itself to which he objects—that cannot be his objection since the second of the two legitimate ways of justifying a priori representations itself depends on a subjective necessity—but rather what he objects to is that on a preformation system the employment of the categories rests only on this subjective necessity. These representations do not “make the experience possible”. Thus, they are “arbitrarily planted in us”. There is no necessary connection between the object and the way we necessarily represent it, and thus we have no grounds for saying that the represented cause is combined with the represented effect, “in the object”.
To recap: what is required for the justification of the employment of a representation is a necessary connection between the object and the representation. Kant sees two ways that this connection might come about. In the case of empirical representations, the object makes possible the representation. In the case of a priori representations, the representation makes the object possible (not regarding its existence, but qua its being the object of representation at all). A preformation system is one in which this necessary connection is severed, and while the representation might happen to correspond to the object, the missing necessity makes a vindication of that representation impossible. Such a system can cite the necessity for the subject to represent the object some way, but not the necessity of the object to correspond to that way of representing it.
What I hope is clear at this point is that inferentialism does not amount to such a system. While it does cite the necessary representative constitution of the experiencing subject in accounting for the validity of the categories, it does not cite only that constitution. Rather, as Kant indicates is required, it casts the categories as making possible the object qua an object of representation. Again, the categories specify what forms an intuition must take if it is to play a role in objectively valid judgements, and thereby what forms the objects represented by such intuitions must also necessarily take. With this vindication in hand, it will be worthwhile to reconsider the passages that Gomes cites in raising his original concern. First, here is Kant’s declaration that he will make clear that his proposal in the Critique is not a preformation system:
[T]he perceptive reviewer may not be left with the necessity, certainly unwelcome even to himself, of taking refuge in a preestablished harmony to explain the surprising agreement of appearances with the laws of the understanding, despite their having entirely different sources from the former. (MAN, AA 4:476n)
What would make a preformation tempting would be if the agreement of appearances and the laws of understanding were surprising, i.e. if there were no element of necessity connecting the laws of the understanding (the representative constitution of the experiencing subject) and appearances. As a version of Kant’s “second way”, inferentialism is safe from this worry. As Kant sees it, Hume is not safe in this way because while he cites features of our representative constitution that are supposedly “subjectively necessary”, these are merely subjectively necessary, but do not thereby correspond to any necessary features of the objects we represent. Pre-established harmony, Kant writes, makes things “only subjectively necessary, but objectively merely contingent, placing together, precisely as Hume has it when he calls this mere illusion from custom” (MAN, AA 4:476n).
The key to understanding Kant’s claim here in a way that is compatible with his own answer to the Justification Problem is to notice and take seriously the ‘only’. It will ultimately be a subjectively necessary feature of our representative constitution that accounts for the objective validity of the categories, but it will do so by being more than just subjectively necessary. It will also make possible the object of representation. That same difference between Kant’s answer and Hume’s underlies the final passage that Gomes cites, from the Prolegomena. There, Kant claims that Hume “passes off […] subjective necessity (i.e. habit) for an objective necessity (from insight)” (Prol, AA 4:257–8).
Hume’s subjective necessities (custom and habit) cannot be guaranteed to represent any real feature in the object precisely because there is no connection between what is subjectively necessary and what is objectively necessary. As Kant puts it, custom and habit might make it the case that we cannot but think of the object in some way, but this subjective necessity is of an entirely different kind than that of the categories, and most importantly does not make possible the object of representation. Custom and habit may be physical constraints on how we represent objects, but the categories are constraints on what it is to think of an object at all. That is, what the categories are subjectively necessary for is thinking of objects. Nothing can count as a representation of an object except insofar as it is subject to the categories. This is what makes possible the object of representation. Again, the very idea of objective validity itself is the idea of a judgement that applies a concept to an intuition, which intuition, in order that it can appear in that judgement must be of a certain form, which form is specified by the categories (B142). That is, what a priori representations are subjectively necessary for is making objectively valid judgements, and in that way they make possible the object of representation. Custom and habit, while they may, de facto, structure our thought in various ways, do not stand in any such necessary relation with the objects of our cognition. It is for that reason that they, unlike the categories, can amount to nothing more than a preformation system. The mistake in equating the two amounts, in part, to misunderstanding the nature of the subjective necessity involved in each case.
3. Jankowiak on Sensations
As I had hoped that he would, Jankowiak focuses his attention on a host of issues surrounding the nature and structure of sensations. Happily, Jankowiak and I agree about a number of points regarding sensations, although sadly I think we actually disagree about the minor point about which we were supposed to fist-bump! I hold that qua the material constituents of an intuition, individual sensations do represent the individual parts of the object represented by that intuition (in the extended sense in which sensations represent at all, that is). To make up for that loss, and to set the stage for our disagreement on more important points, here is a brief catalogue of other fist-bump-worthy points of agreement. Jankowiak and I agree that sensations are the material constituents of intuitions. We also agree that whereas intuitions represent complex objects as complex, sensations do not (although they do serve a representative role nonetheless). We also agree that the sensations that are the matter of representations of objects of outer sense do not represent mental states, but rather the parts of those objects. (I additionally hold that a kind of sensation is necessary to account for the matter of our representations of inner sense [see my book, pp. 270–1], but I don’t know if Jankowiak holds that too.) Finally, and by way of transition to the first of Jankowiak’s concerns that I want to address, we agree that the representation of spatial and/or temporal complexes must have both a matter and a form that is not merely conceptual.
That final point of agreement seems to be one that Jankowiak will find surprising, as he worries that I have committed to a position according to which “the form of mental representation is conceptual rather than spatial”. Jankowiak rightly chastens me for saying “very little about space as a form of intuition”, and my only defence is that in fighting a battle simultaneously on two fronts, I focused my attention on the more pressing enemy—the dreaded non-conceptualist who holds that representations of spatiotemporal complexes are entirely non-conceptual—to the neglect of the one that I had presumed had long since vanquished—the radical conceptualist that Jankowiak takes me to be. To put it another way, here is one crude way of divvying up this philosophical space:
(a) Those who hold that representations of spatio-temporal complexes as complex are entirely conceptual (radical conceptualists, including Leibniz).
(b) Those who hold that representations of spatio-temporal complexes as complex are entirely non-conceptual (non-conceptualists, including, e.g. Hanna and Allais).
(c) Those who hold that representations of spatio-temporal complexes as complex necessarily have both conceptual and non-conceptual aspects (me).
In Kant’s Inferentialism, I focused my attention on rebutting those scholars in group (b), who have been gaining in prominence for some time. Jankowiak is right, however, that having done so, it is all too easy to read me as having joined the ranks of those in (a), and making it clear that I have not is a debt that needs paying. In doing so, though, I also hope to make clear why my way of avoiding radical non-conceptualism is not Jankowiak’s, and so to bring out one of the aforementioned important points of disagreement.
The worry, then, is that I have committed myself to the view that the only structure that sensations exhibit in an intuition is a conceptual one. Here is one reason for thinking that that cannot be right. As I understand it, in most fundamental form, an intuition makes singular determinate reference to an object as the necessary connection of its parts. It is as uncontroversial as it gets, I think, that Kant holds that doing this cannot be a purely conceptual affair. Consider, for example, the argument from incongruent counterparts. Roughly, the idea here is that Leibniz had held that one (God) can secure singular determinate reference to an object (monad) via a complete conceptual description of that object. Kant leverages the fact that a concept is, by definition, a kind of representation that can apply to many distinct objects to demonstrate that there is no in-principle reason to think that there could not be qualitatively identical but numerically distinct objects. He also then cites what he takes to be actual examples of such objects to show that, using Leibniz’s own principle of the identity of indiscernables, objects of representation must have properties that are not represented by concepts alone. That is to say, objects must have spatiotemporal properties. Now, Jankowiak does mention that I hold that sensations are themselves non-conceptual components of conceptually-structured intuitions, and so might hold that that is sufficient for securing the necessary non-conceptuality of the representation of spatiotemporal properties. I agree with him, however, that this is not enough. In addition to having sensations as their non-conceptual components (matter), intuitions must also have a non-conceptual structure (form) in order to make determinate singular reference to spatio-temporally located objects. Again, Jankowiak and I agree on that.
Now for the disagreement. Jankowiak’s proposal is that—perhaps in addition to conceptual form, or perhaps in place of it—sensations must be understood as also having a distinctly spatial form:
Moreover, Kant also indicates in multiple places that the organisation given to sensations (which, again, are the matter of intuition) is spatial. He says that sensations are “ordered and placed” (A20/B34) in space and that sensations are represented “outside and next to one another” (A23/B38), which is a pretty clear indication that the form given to collections of sensations is a spatial form.
To be very frank, it seems to me that there are two ways of understanding Jankowiak’s claim that sensations have a spatial structure, one of which I mostly agree with, and the other of which simply cannot be right. Here are two claims that Jankowiak might be making.
(a) What sensations represent they represent as spatial.
(b) Sensations themselves are spatial arrangements of mental states.
I think that (b) must be wrong, although I admit that it is not obviously so. After all, mental states could turn out to be brain states, and therefore spatial. While that line of thinking may well be right, the important point in understanding Kant on sensation, I think, is that before we can pursue questions of the ultimate constitution of mental states, we must understand the representative role that those states play in our cognitive lives. Even if intuitions turn out to be spatial arrangements of sensations, they do not play the representative role that they do (representing objects) in virtue of being such an arrangement. As Jankowiak points out, sensations (and intuitions) are by definition mental states, and are therefore possible objects of inner sense. (More on that in a moment.) Kant is very clear, though, that the form of inner sense is time alone. That is to say, in representing mental states as mental states, we represent them as existing in time, but not in space. Thus, it cannot be their spatial form in virtue of which intuitions represent what they do.
Ok, so what about (a), the claim that what sensations represent, they represent as spatial? If we limit ourselves to objects of outer sense, I take it that as an interpretation of Kant, this claim ought to be trivial. What intuitions of outer sense, composed of sensations, represent, is represented as spatial. Jankowiak and I certainly agree on that much, but the question, as I understand it, is in virtue of what feature of an intuition it represents its object as spatial. That is to say, what about an intuition makes it such that it represents a spatial complex as complex? In my book, I argue that one necessary condition of any representation of a complex states of affairs or complex object is that it be conceptually structured. Jankowiak rightly notices that this necessary condition is not also sufficient. Concepts alone cannot represent spatiality. So what more is needed? Well, this brings us to Jankowiak’s second main worry about my account.
One easy way to begin to answer the question of what the internal structure of an intuition is in virtue of which it represents a spatial complex would be to simply introspect and observe what structures intuitions in fact have. As Jankowiak observes, however, that answer is not available to me because I hold that sensations and their structure are theoretical-explanatory, or transcendental, posits. What we represent are first and foremost objects as the necessary connection of their parts. We can also introspect and represent our own mental states (as they appear to us). In asking after the features of our mental states in virtue of which we form such representations, however, I do not think that we are asking after any of their phenomenal qualities (at least not as such). Rather, this is another form that Kant’s “how possible” question takes, and as such its answer is not empirical, but transcendental. It asks what cognition must be like for creatures like us, if we are to have such representations, and an appropriate answer to that question gives the necessary conditions of cognition, which are not themselves available to mere observation. Below are a few of the texts in which Kant explicates what he means be sensation. Notice that in all of them, sensations are defined by the representative role they play in cognition, rather than by an appeal to introspective ostension.
The effect of an object on the capacity for representation, insofar as we are affected by it, is sensation. (A20/B35)
The genus is representation in general (repraesentatio). Under it stands the representation with consciousness (perceptio). A perception that refers to the subject as a modification of its state is a sensation (sensatio). (A320/B377)
The primary elements of our cognitions are sensations. This is what one calls those representations in which the mind is regarded as merely passive, acted upon by the presence of an object. (Refl 619, AA 15:268)
Kant repeatedly defines ‘sensation’ in functional terms. Sensations are the effect of an object on the capacity for representation; they are modifications of the state of the subject; they are what one calls those representations in which the mind is regarded as merely passive. What I argue in Kant’s Inferentalism is that it is no mere accident that Kant does not give any intrinsic characterisation of sensations. Rather, this is because he posits sensations, the non-conceptual causes and components of our conceptually-structured intuitions, as a way of explaining the possibility of those intuitions themselves.
Thus, because I take sensations to be theoretical-explanatory transcendental posits, there is only so much that I can say about their structure. I hold that what we know about their structure, we do not know through observation, but only by an analogical extension of the structure of the representations that these are meant to explain. That is, since sensations are posited as a means of explaining our conceptually-structured representations of spatial and temporal complexes, we can conclude that they are modifications of the states of the experiencing subject that have a space-like and time-like structure—Sellars calls them σ- and τ- manifolds—as well as a structure that is isomorphic to the structure of our judgement-level inferential practices, but can say little else about them. The internal structure of an intuition is ex hypothesi a sui generis structure.
Which brings me to one final point of clarification regarding my claim that the structure internal to an intuition is a conceptual structure. As Jankowiak points out, this claim appears to be in tension with what is perhaps the most central thesis of my book, namely, that concepts are inferential rules. Roughly, we picture necessary connections between worldly objects by licensing inferences between judgements containing representations of these objects (intuitions). To repeat an example, one might be inclined to take the following argument to be an enthymeme:
1. Suppose there were an elephant’s tail in front of me now.
2. There would be an elephant’s body at the end of that tail.
If one were so inclined, then the premise that would appear to be missing here would be something like,
3. If there is an elephant’s tail in front of me, then necessarily there is an elephant’s body at the end of that tail.
It is because the argument concerns subjunctive conditionals that this additional premise must be one about what is necessary, and not merely actual. What this strong modal claim implies, though, is that it is impossible for its antecedent (the equivalent of 1) to be true and its consequent (the equivalent of 2) to be false. That, however, amounts to the same thing as the original argument’s being valid. Of course, it is not valid in virtue of its logical form, but rather in virtue of the content of the concepts deployed in the judgements involved. Thus, representing the necessary connection between two worldly objects can be achieved by licensing a certain inference between judgements containing representations of those parts, and it is the concept ‘elephant’ that plays this dual role: it represents an object-level necessary connection by serving as a meta-level inferential rule.
The tension that Jankowiak notices stems from the fact that since sensations do not appear in judgements, they cannot be linked via inferences in the way that intuitions are, and so appear to be incapable of exhibiting an inferential, or conceptual, structure. I’m grateful to Jankowiak for drawing attention to this issue, which I address in Chapter 3, but the discussion of which is admittedly one of the more difficult ones in an already difficult book! I also applaud Jankowiak for getting the form of the response I give to this problem exactly right! I take as my jumping off point Kant’s claim that
[s]ynthesis in general is, as we shall subsequently see, the mere effect of the imagination, of a blind though indispensable function of the soul, without which we would have no cognition at all, but of which we are seldom even conscious. (A78/B104)
I argue that what Kant means by this is that the synthesis that the imagination affects internal to an intuition does not operate by consulting a concept-qua-inferential-rule, then comparing the conditions of application described by this rule with a sensation to determine whether that sensation meets those conditions, and then applying the rule to those sensations. Rather, the sense in which the imagination is blind is precisely that it carries out this synthesis by following a concepts-qua-inferential-rules without consulting them. On the one hand, the imagination imparts a structure to sensations that is isomorphic to the inferential structure in which intuitions stand, but does so without having to represent those sensations as subject to those rules. On the other hand, though, it also does not do so “merely by accident”. The imagination is, in some sense, trained by the understanding to do exactly this, and thus non-accidentally produces intuitions with the forms and content required by the understanding. So, one way to put my interpretative claim here is that the internal structure of an intuition, while not strictly-speaking conceptual, is nonetheless necessarily in accordance with concepts, and non-accidentally so. They have the structure of concepts imposed on them, by the blind imagination, even if they are not themselves subject to concepts-qua-inferential rules.
Jankowiak describes this interpretation as being “that intuitions have a structure that is not itself inferential, but which results from an inferentially-guided process”, which is a description that I would accept. He then objects that,
[t]o use an analogy, if I put together a bunch of Lego bricks in accordance with the concept ‘house’, there’s no reason to think that the resulting building is itself in any way conceptual or inferentially structured.
Jankowiak might well be right that it is inappropriate to call the structure internal to an intuition that is at issue conceptual or inferential. I would be fine with conceding as much because the claim that I really need to defend isn’t that such structures are conceptual or inferential, but rather that they are structured or governed by conceptual norms, and I think I still get that. Let me explain. As Jankowiak helpfully summarises, in Chapter 2 of my book I argue that Hume’s theory of mental representation suffers from a determinacy problem: the laws of association that structure Humean representations are not sufficient for accounting for the determinate content that Hume himself would agree that our representations have. I won’t rehearse the argument for that again, but suffice it to say that as I read him Kant is particularly concerned to address this shortcoming of Hume’s theory, and does so in part by taking the relations that structure our representations to be essentially normative, and more specifically to be norms of inference (see above). Thus, I take Kant to be committed to the thesis:
(a) All representations of complex states of affairs as complex are structured by conceptual norms.
I also take him to be committed to the claim:
(b) Intuitions represent complex states of affairs as complex.
Finally, if I accept Jankowiak’s correction above, I should also take him to be committed to the claim:
(c) Intuitions are not conceptually structured.
Prima facie, those three claims might appear to constitute an inconsistent triad, but I do not believe that they do. Specifically, I take my understanding of the structure internal to an intuition, and its formation by the blind imagination, to be a way of showing that something can be subject to conceptual norms without having those norms applied to it directly, or as Jankowiak would put it, without itself being conceptually structured. The Lego house may not be conceptually structured, but its construction, and its being properly subsumable under the concept ‘house’, nonetheless require it to be subject to the concepts-qua-inferential-rules that constitute that concept, if not as an ought-to-do rule, then as an ought-to-be rule.
Received: 1 January 2017.
 As we will see in a moment, Kant repeats this way of presenting his view almost exactly in introducing preformation systems at B167.↩
 It is worth noting that I take this to be merely the fundamental form that an intuition takes, but that I also hold that it can take other forms. I.e. since we can make judgments with universal, particular, or singular form, we must be able to form intuitions of unities, pluralities, and totalities. Jankowiak disagrees, holding that intuitions can only make determinate singular reference. I infer that Jankowiak holds that the categories, which of course correspond to the logical forms of judgments, do not represent the conceptual forms of intuition. Here is one place where Kant appears to take my side. It is in a note that accompanies the table of the categories in Kant’s copy of the first edition:
Logical functions are only forms for the relation of concepts in thinking. Categories are concepts, through which certain intuitions are determined in regard to the synthetic unity of their consciousness as contained under these functions; e.g., what must be thought as subject and not as predicate. (HN, AA 23:25)
Intuitions represent objects, substances, or that which must be thought as a subject. The logical forms of judgement contain the forms that any such subject can take, in this case their various quantities: This S is P; Some S are P; All S are P. The categories are the concepts of these forms of intuition. What is represented by ‘This S’ is a unitary object; by ‘Some S’ a plurality of objects; by ‘All S’ a totality of objects. It can certainly be argued that singular judgements do occupy an especially fundamental place among these forms, given their connection with perception, but it seems clear to me that this is not the only form that an intuition can take.↩
 For a more thorough discussion of this argument, specifically in the context of the illicit use that non-conceptualists make of it, see Landy (2013).↩
 Just to be clear, I hold that intuitions must also have a conceptual structure, since they are complex representations of complex objects, and all such representations have conceptual structure. Thus, it follows that intuitions must exhibit multiple different kinds of structure. I do not see that as a problem, but thought it best to reiterate the commitment, which I discuss at more length in the ‘Postscript on Transcendental Idealism’ of my book.↩
 More precisely, I argue that any representation of a complex state of affairs or complex object must be structured by conceptual norms. This difference will make a difference in addressing Jankowiak’s concerns about the structure internal to an intuition in a moment.↩
 That said, qua mental states, it must also be possible to accompany sensations with the ‘I think’. As McDowell (of all people!) points out, this means that sensations actually play both a transcendental and an empirical role. In their role as a theoretical posit that explains our cognition of objects they are merely transcendental, but as the object of an act of introspection, they cease to play that explanatory role, and instead become a merely intuited item, which however is represented only as it appears, and not as it is in itself. It is because of that final clause that this provides no succour to Jankowiak. See McDowell (2009:447), Rosenberg (2007:277), and my Kant’s Inferentialism, ch. 5.↩
Landy, D. (2013), ‘What Incongruent Counterparts Show’, European Journal of Philosophy 21(4): 507–24.
McDowell, J. (2009), Having the World In View (Cambridge, MA: Harvard University Press).
Rosenberg, J. (2007), Wilfrid Sellars: Fusing the Images (Oxford: Oxford University Press).
© David Landy, 2017.
David Landy is an Associate Professor of Philosophy at San Francisco State University. He works primarily on the history of modern philosophy, especially Hume and Kant, and also has interests in German Idealism and the work of Wilfrid Sellars. He has published in, among other places, the EUROPEAN JOURNAL OF PHILOSOPHY, HISTORY OF PHILOSOPHY QUARTERLY, INQUIRY, JOURNAL OF THE HISTORY OF PHILOSOPHY and KANTIAN REVIEW.