By Tim Jankowiak
One of the central interpretative claims of David Landy’s Kant’s Inferentialism: The Case Against Hume is that “both Hume and Kant are first and foremost engaged in projects aimed at giving a philosophical account of the nature of mental representations” (p. 2). This approach to reading Kant (and Hume) colours the entirety of Landy’s book, from his initial accounts of Kant’s response to Hume, through his readings of the Deductions, the Analogies, and the Paralogisms, all the way, finally, to his Sellarsian appraisal of transcendental idealism. Regarding the latter, he denies that transcendental idealism
is a claim about what we can know about noumenal objects, or a claim about what noumenal objects must or must not be. [Rather,] Transcendental Idealism is primarily a thesis about what can and cannot be represented. (277)
I’ll state at the outset that I whole-heartedly agree with Landy on this point. This is not to say that Kant is not engaged in metaphysics and epistemology too; both are surely part of his project. But the metaphysical and epistemological results of the Critique of Pure Reason are contingent upon and determined by his results about the structures and limitations of mental representation in humans.
Landy’s reading of Kant’s theory of mental representation will be the focal point of this essay. I shall be most concerned to evaluate the details of Landy’s understanding of a Kantian intuition. Landy is an inferentialist conceptualist, arguing that intuitions should be understood as conceptually structured sensory pictures of the world. While I do think he makes a good case for a broadly inferentialist reading of Kant’s theory of representation, I run into some trouble with the details of his claims about the nature of intuitions. Specifically, I’ll call into question Landy’s claim that intuitions should be understood as collections of sensations given conceptual/inferential form.
In the following, I’ll first briefly recount Landy’s analysis of Hume’s theory of representation (against which Kant’s will be contrasted). In the second section, I reconstruct Landy’s interpretation of Kant’s theory of sensation and intuition. And in the final section I offer some criticisms of his interpretation.
1. Hume’s Theory of Mental Representation and the Problems with it
To understand Landy’s reading of Kant’s theory of mental representation (and of intuition specifically), we must begin with his take on Hume’s theory of mental representation, which will function as a foil for Kant’s. As Landy sees it, both Hume and Kant defend a sort of ‘picture theory’ of mental representation: both argue that the basic units of mental representations are mental ‘pictures’ of worldly states of affairs. Where they differ is in their takes on the components of these representations, and how they’re structured.
The principal target of Landy’s analysis is the Humean ‘idea’. It is already well-known that Hume ties the representationality of an idea to its counterpart, the ‘impression’: there can be no idea without a corresponding impression. It is also well-known that the relation between idea and impression is one of strict resemblance: ideas are copies of impressions (one thing is a copy of another so long as the first was caused by and exactly resembles the second). This leads Landy (after some engagement with other interpretations in the literature; see pp. 22–33) to propose a Humean Representational Copy Principle:
Representational Copy Principle: “the representational content of any perception is that of which it is a copy”. (p. 29)
The Representational Copy Principle is initially supposed to account for the representationality of simple ideas. Landy’s real concern, though, has to do with complex representations that represent complex states of affairs. On Landy’s reading of Hume’s pictorial theory of representation, a collection of ideas can represent a complex state of affairs when the ideas are arranged in the same way that the objects represented are arranged. This requires thinking of ideas as the sorts of things that can be given spatial determinations (since we are frequently representing spatially arranged states of affairs).
We represent a as being next to b by placing an idea of a next to an idea of b. The idea of a spatial complex is nothing more than a spatial complex of ideas. (p. 45)
More generally, the Humean schema for a representation of complexes as complex looks like this: If simple idea ‘x’ represents object x (and ‘y’ represents y), and if x and y stand in relation R, then ‘x’R‘y’ represents xRy. Thus we see that Hume’s theory of mental representation appeals to copying not just at the level of simples, but of complexes as well.
Landy thinks that Kant has the resources to refute the Humean theory of representation. Before turning to this refutation, I’ll just mention that I’m not convinced that Kant actually intended to aim these arguments at Hume. The passages that Landy cites in his reconstruction of these arguments just don’t seem to be places where Kant had Hume in mind (he relies heavily on Reflexionen and logic lecture transcripts, as well as some passages from the Critique). At the very least, Kant doesn’t explicitly mention Hume in these passages. Nevertheless, whether these are arguments that Kant did aim at Hume or instead merely arguments that he could have aimed at Hume, I think Landy does make a strong case that Kant has at least the resources to run such anti-Humean arguments.
Landy finds in Kant two arguments against the Humean theory of representation. I’ll focus only on the first of these, which accuses Hume of mistakenly allowing for a sort of representational overflow: Hume’s representations represent too much, and often end up representing the wrong sort of thing. Since all that’s required for a certain feature of a state of affairs to be represented is that it be mirrored in the idea, it follows that every idea will represent a great deal of content, and there is no way to single out any one particular feature as the actual target of the representation.
Take temporal representations. If a representation represents a preceding b, it is because the representation involves an idea of a followed by an idea of b. The problem with this is that any representation that involves a sequence of perceptions will thereby be a representation of succession. Kant’s well-known examples (from the Second Analogy) of the perception of the parts of a house and the perception of a ship sailing down-stream become relevant here. In both cases there is a succession of perceptions. Clearly though, only one of these successions of perceptions should also be considered a perception of a succession. But if Hume’s theory of representation were true, then since I perceive the eastern side of the house after the southern side of the house, it must be the case that I am representing the eastern side of the house happening after the southern side, which is nuts.
Another example: If I perceive a penny next to a dime, my representation will be a representation of all of the following contents: the penny being larger than the dime; the dime being smaller than the penny; the penny being darker than the dime; the dime being brighter than the penny; the dime being more valuable than the penny; the penny less valuable than the dime; the penny being heavier; the dime being lighter; the two being worth 11 cents; the difference between the two being 9 cents; the quantity 2; and so on for innumerably many more contents. My representation will be overflowing with this abundance of contents even if I only glanced at the coins for a second while wondering how much change was in my pocket. It would seem that my attention was on only the value of the coins. But if Hume were right, then my representation of their value would be no more or less salient than any of these other contents. This is simply not phenomenologically plausible.
Landy does consider a possible Humean response that appeals to Hume’s theory of associations, but this too turns out to be unworkable. The conclusion of the discussion is that Hume lacks the resources to explain how a pictorial representation can focus on specific features of the represented state of affairs while ignoring others. This is where Landy’s Kant marks an improvement over his Hume. As I’ll explain in the next section, Landy argues that mental pictures are not structured as mere copies of the relations among the represented state of affairs; rather, they are structured by ‘concepts-qua-inferential-rules’. Since only some of the relations among the represented objects will find corresponding concepts within the representation, it will be possible to explain how we can attend to some features of objects at the expense of others.
Before turning to Kant, I’ll just mention one other problem with Hume’s theory of representation, which is perhaps at least implicit in Landy’s critique. If the Representational Copy Principle really does reflect Hume’s considered opinion about where representationality comes from, then I worry that we lose track of what representations really are. Again, as Landy reads Hume, x represents y if and only if x exactly resembles and is caused by y. Although Hume takes it for granted that x will be an idea, there’s nothing in the theory that requires this. Rather, it seems that anything that meets these two conditions (exact resemblance and causal connection) will count as a representation on this theory. Thus, a photograph can exactly resemble the scene it depicts and is caused by that scene, hence it represents in the exact same way that an idea of the scene would represent it. If I were to make a mould of my hand, the replica would resemble my hand and have been caused by it, hence it would represent my hand in the same way that an idea represents my hand.
This is not to say that the photograph or the replica of my hand would be elevated to the status of mental states. Rather, the problem is that ideas have had their status downgraded: we’ve lost track of the sense in which the representationality or intentionality of an idea is a mental phenomenon. That is, we haven’t imbued all copies with the mental; we’ve taken the mental out of ideas. Perhaps this is what Hume intended, and it’s just one more facet of his attempt at a naturalised understanding of the mind. Perhaps it reflects a misunderstanding on Landy’s part (though I don’t claim to have thought about this aspect of Hume’s theory as much as Landy has). Either way, it strikes me as problematic, and so if this really is Hume’s theory of mental representation, then that theory seems deeply flawed. Once again, Kant will have a more satisfying story to tell here. Since representation, for him, depends on conceptually-guided synthesis, and since synthesis must be performed by an apperceptive understanding, the sense in which representations are distinctly mental will be preserved.
2. Kant’s Theory of Mental Representation
Understanding Kant’s theory of mental representation depends first and foremost on an interpretation of Kant’s theory of intuition. Landy offers a ‘broadly Sellarsian’ (that is, inferentialist) reading of Kant’s theory of intuition. He argues that intuitions should be construed as collections of sensations organised and related by concepts-qua-inferential-rules. In this section, I explain Landy’s interpretation, and in the next I criticise a few aspects of it.
First off, Landy argues that intuitions are essentially complex representations of complexes as complex. That is, intuitions have components and they represent objects or states of affairs that themselves have various components, and they represent these objects or states of affairs as having these components. One might worry that this is a violation of Kant’s claim that intuitions are strictly singular representations. Although I do not recall Landy himself addressing this point, I take it that his claim is consistent with the singularity criterion. For even if the objects of intuition have parts, these objects are still singular and individual objects.
So both Hume and Kant have accounts of representations of complexes as complex. Where Hume thinks that a complex representation represents a complex object or state of affairs by mirroring the very same relations that are in the object or state of affairs, Kant has a ‘counterpart theory’ of representation. Landy finds motivation for this reading in an analogy found in Kant’s notes on Meier’s Auszug aus der Vernunftlehre:
What is it then in the representation that is in agreement with the represented things? Since the representation borrows its ground from the represented thing, it agrees with the latter in that it is composed out of its partial concepts in the same way that the whole represented thing is composed out of its parts. E.g., one can say that the notes of a musical piece are a representation of the harmonic combination of the tones—not as if a note were similar to a tone, but because the notes themselves have a combination among themselves like that of the tones themselves. (Refl 1676, AA 16:78; quoted by Landy at p. 65)
The notes in a score represent tones, and the relations between the notes represent the relations between the tones. Written notes are related by being placed higher and lower on the staff, and the tones themselves are related in terms of their pitch. Although these are different relations, since changes in the one correspond systematically with changes in the other, the written notes can function as representations of the music itself.
This leads to Landy’s basic point of contrast between Hume’s and Kant’s fundamental schemas of mental representation. Where Hume claims that ‘x’R‘y’ represents xRy, Kant thinks that ‘x’R*‘y’ represents xRy (where R* is a relation which is distinct from, yet a “counterpart” to R).
So much for the generic schema of representation in Kant. We now want to know, specifically, what the relata are in Kantian representations (the ‘x’s and ‘y’s), and what the counterpart relations are (Landy’s R*). The short answer is that the components of the representation are sensations, and these sensations are structured conceptually. But there’s more to say here. Let’s take sensations first.
Kant can be frustratingly ambiguous about the sense in which sensations are representations, or even whether they are representational at all. Sometimes he suggests that sensations are merely the causal products of the affection of objects on the senses. Sensations are produced by objects affecting the senses, and thereby these resulting mental states ‘correspond’ to the part of the object that is causally responsible for the sensation (cf. A20/B34). Of course (pace Hume), the fact that a mental state was caused by an object is not yet sufficient to entail that the mental state also represents anything. It may just be a brute, non-indicative mental occurrence. Accordingly, many have understood Kant to hold that sensations are not really representations at all (see Pippin 1982:31; McDowell 1996; Allais 2009:398).
Other times though, Kant does attribute a representational function to sensation. Sometimes he indicates that sensations are “subjective representations” (A28/B44) that represent the modified state of the subject: “A perception that refers to the subject as a modification of its state is a sensation” (A320/B377). Some have taken this to be Kant’s considered view (see Aquila 1983:59; Watkins 2012:321; Thompson 1972:323).
And at still other times he indicates that sensations represent not the subject itself, but the external objects which caused them: “[S]ensation is that which designates an actuality in space” (A374); the matter in an empirical object is the “object of sensation” (B207). This is Landy’s view (and mine; see Jankowiak 2014).
So we have three claims about sensations:
(1) They are caused by the affection of objects on the senses;
(2) they represent the modified state of the subject;
(3) they represent the matter in external objects.
My own view is that (2) and (3) are incompatible. I think that Kant is being misleading when he occasionally suggests that sensations only represent internally. I take his considered view to be that sensations typically function as representations of objects, not the subject (more on this shortly). Landy doesn’t seems to notice the tension, and he attributes both claims to Kant in quick succession (see pp. 156–7). Either way, Landy’s considered view is that, whatever else, sensations are definitely representations of external objects. He says that “the object represented is represented as a complex of parts via a representation that is itself a structured collection of sensations, which are themselves representations of these parts” (p. 132). The question then is what kind of representations they are.
Landy makes a distinction between two stages at which sensations appear in the perceptual/cognitive process. First, there are ‘pre-perceptual’ sensations, i.e. sensations insofar as they have been caused by affecting objects and so correspond to parts of the object, but which are not yet available to consciousness and hence are not yet representations of anything (for correspondence to an object is not yet representation of the object). The occurrence of these pre-perceptual sensations then prompts the imagination into action, causing the formation of intuitions. Thus secondly, these sensations can be “taken up” in perceptual synthesis (as described in the A-Deduction; see Landy, pp. 132–41) and formed into intuitions: “Kant takes sensations to be not just the causal antecedents of intuitions but also their constituents” (p. 159). Once sensations are put to work as the components (the “matter”) of intuitions, they can represent. (I also agree with him on this point.)
It is a bit difficult to pin down the precise sense in which sensations are representational for Landy. He insists that sensations are not conceptually structured (collections of sensations are—more on that later—but individual sensations are not). Since Landy correctly insists that a representation cannot represent its object as anything unless it is conceptually structured, it follows that individual sensations will not represent anything as anything. So how do sensations represent? Landy appeals to the Sellarsian distinction between representations that “stand for worldly objects” and those that “stand in for worldly objects” (p. 156). The latter of these relations—’standing in for’—characterises the representationality of sensation. Sensations function as “mental proxies” (ibid.; Landy borrows the phrase from Sellars). “They represent the worldly objects that they do, not in the sense that a picture represents a scene, but rather in the sense that a lawyer represents his client” (ibid.).
While I don’t find the analogy with the lawyer particularly illuminating, I think I can gather what is going on here. Representations that ‘stand for’ objects indicate or refer to objects distinct from themselves. Sensations don’t do this. Representations that ‘stand in for’ objects do not refer, but are rather themselves referred to in other representational states. And by being systematically related to external objects, they enable the representation of those external objects. I’ve elsewhere (Jankowiak 2014) explained a similar point by appeal to an analogy with the pixels on a computer screen. If I see a live video feed of my friend, I take myself to be seeing her, and I ignore the fact that I am looking at an array of illuminated pixels. There is no sense in which each individual pixel refers, for instance, to a part of my friend’s face. However, they do ‘stand in for’ the parts of her face because their colours are systematically related to the colours and contours of her face, and in attending to those pixels I am able to take myself to be looking at my friend. This is how I think sensations represent, and I think it’s Landy’s view too (if so, imagine David and I are fist-bumping here). And to continue with the analogy one step further, note that any one pixel taken out of context would not be able to ‘stand in for’ anything. Rather, it is only in the context of a massive array of pixels that any of them can stand in for the parts of my friend’s face. This corresponds to Landy’s point that sensations prior to their synthesis into intuitions don’t represent at all.
So much for sensations, i.e. the relata in Landy’s schematic for Kantian representations (‘x’R*‘y’ represents xRy). What about the relations, i.e. Landy’s R*? Landy holds that intuitions are conceptually structured collections of sensations. Hence, he is firmly in the conceptualist camp. However, he is a moderate conceptualist insofar as he admits that intuitions, while being essentially and necessarily conceptual, nevertheless have a non-conceptual component (i.e. the individual sensations).
One of Landy’s most basic claims is that “all representations of complexes as complex are conceptually structured” (p. 121). Citing the Metaphysical Deduction (see A77/B102–3), he reminds us that while receptivity can produce a manifold, it will only be possible to comprehend the manifoldness of the manifold in one unified representation if synthesis is involved. And he identifies all synthetic activity with the function of a concept (notice that this move is controversial, as the non-conceptualist will deny that all synthesis is conceptual). Hence, “all representations of a manifold as a manifold (a complex as a complex) are produced by the imposition of a conceptual structure on the manifold” (p. 122). Since intuitions are representations of complexes as complex, it follows that they are conceptually structured. Additionally, he claims that the concepts that can provide structure to collections of sensations in intuition can also structure the relations among several intuitions together. (In fact, it sometimes sounds as though the full ‘pictures’ that Landy is describing are not individual intuitions themselves, but rather conceptually structured collections of intuitions, which are themselves conceptually structured collections of sensations; see p. 127.)
What is peculiar about Landy’s reconstruction is not so much the conceptualist line of interpretation itself (though that is a contentious position in itself), but the specific way in which concepts are supposed to be at work in intuitions. He argues that concepts should be understood as inferential rules. That is, any one concept, when at work in a representation of an object, will require, license, or prohibit various other conceptualisations. (E.g., if I represent something as a body, then I must represent it as extended, I may represent it as a stone, and I may not represent it as a soul.) This Sellarsian/Brandomian take on Kantian conceptualism is at the heart of Landy’s entire book, and his conclusions regarding the arguments of the Analogies (Chapter 5) and the Paralogisms (Chapter 6) rely heavily on this inferentialist interpretation of intuition. While I am attracted to many aspects of Landy’s reading, there are others that I find troubling. I turn now to some of these difficulties.
3. Some Concerns about Landy’s ‘Pictures’
One of the things that has puzzled me most about Landy’s reading of Kant’s theory of mental representation is his claim that Kant, like Hume, defends a ‘picture theory’ of mental representation (though Kant’s pictures take a very different form than Hume’s). His claim is that these pictures are composed of sensations which are related to each other via concepts-qua-inferential-rules and which thereby constitute intuitions; multiple intuitions can then be related to each other by further concepts-qua-inferential-rules. Where sensations are the matter of these mental pictures, concepts are the form. For the life of me, I just can’t wrap my head around the notion of a ‘picture’ whose structure is conceptual. If anything, it would seem to be more natural to contrast conceptual representation with pictorial representation.
My thought would be that the structure or form of a picture would be some sort of spatial organisation. After all, spatial organisation (at least in two dimensions) is the form of actual pictures. Moreover, Kant himself explicitly and repeatedly states that space (and time) are the forms of intuition. For instance, when first introducing the notion of an intuition in the Transcendental Aesthetic, he tries to isolate the form of intuition by abstracting away both sensory matter and the understanding’s concepts:
If I separate from the representation of a body that which the understanding thinks about it, such as substance, force, divisibility, etc., as well as that which belongs to sensation, such as impenetrability, hardness, color, etc., something from this empirical intuition is still left for me, namely extension and form. (A20–1/B35)
The same point, a moment later:
In the transcendental aesthetic we will therefore first isolate sensibility by separating off everything that the understanding thinks through its concepts, so that nothing but empirical intuition remains. Second, we will then detach from the latter everything that belongs to sensation, so that nothing remains except pure intuition and the mere form of appearance, which is the only thing that sensibility can make available a priori. In this investigation it will be found that there are two pure forms of sensible intuition as principles of a priori intuition, namely space and time. (A22/B36)
Kant makes it clear in these passages and others that the form of intuition is space (and time), and he even explicitly states that we must abstract from understanding and its concepts in order to isolate the form of intuition. 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. All this makes it rather peculiar that Landy says very little about space as a form of intuition. (There are discussions of intuitions which are representations of space at pp. 56–60 and 122–7, but not of space as the form of those representations.)
If the form of mental representation is conceptual rather than spatial, then I worry that appealing to the metaphor of a ‘picture’ stretches that metaphor beyond the point of usefulness. It seems that if a conceptually-structured, non-spatial representation can count as a picture, then any old representation might as well be called a picture too. And if that’s the case, then explaining Kant’s theory by appeal to the notion of a picture doesn’t end up illuminating very much.
Things are even worse when we note that, for Landy (and Sellars before him), “sensations are posited on broadly explanatory grounds” (p. 293). The idea is that sensations are not something of which we are directly aware (hence, I take it, not something that can be phenomenologically available to the perceiving subject). Rather, we infer the existence of sensations from the fact that we have empirical representations and that there must be some sensory stimuli to explain why we have these representations instead of others. This is the “sense impression inference” (p. 294). Further, the qualities possessed by sensations, and the structures they instantiate, must be “analogous” to, but ultimately distinct from, the qualities and structures that we end up representing.
We must suppose that sensations have a structure that is analogous to that of space […]. Likewise we can stipulate that sensations must have a structure that is analogous to that of time. (p. 293)
Sensations must have characteristics that are not colors but that are sufficiently analogous to colors to explain why it is that we respond conceptually to colored worldly items as we do. (p. 294)
Because these structures and qualities are beyond the range of conscious, phenomenal access, we cannot know what they are, and can discuss them only obliquely as the structures and qualities that are not space, time, colour, etc., but are inferred (through the sense-impression inference) to be in some sort of isomorphic relation to space, time, colour, etc. I’ve always found the Sellarsian claim about the ‘sense-impression inference’ a bit odd, because it seems that most people do enjoy a pre-theoretical awareness of their sensations as sensations. But I find the idea especially perplexing in the present context, due to the function that Landy has assigned to sensation: Sensations ‘stand in for’ their objects. As I understand Landy (perhaps I am mistaken here), this means that sensations are (in some sense) objects of awareness, but are not themselves intentionally directed at anything else. I do not see how sensations could perform this function if their very existence is not available to normal conscious perception and if we only come to believe in them on the basis of a philosophical argument.
Anyway, here’s the story as I understand it: Intuitions (and groups of intuitions) are pictures of objects. The components of these pictures do not resemble the parts of the objects that they correspond to (except insofar as isomorphism is a resemblance, but that would be stretching things, I think). Further, the components of these pictures (sensations) are not even phenomenologically available, as we only believe that these components exist at all on the basis of a sense-impression inference. Thus we cannot ‘see’ the components of this picture; and if we cannot see any of the components of the picture, then we cannot see the picture itself. That is, the picture itself is not phenomenologically available to the perceiving subject. And then on top of all of that, the picture has a conceptual (but apparently not spatial) form.
With all these aspects of Landy’s theory taken together, I lose track of the sense in which these representations can be called ‘pictures’ at all. This is too bad, because in calling the theory a ‘picture theory’ at the beginning of the book, I was expecting a phenomenologically grounded account of mental representation, but the theory instead ends up far removed from the phenomenology of normal human perception.
Perhaps, though, one shouldn’t ask too much of metaphors. So let’s put aside the question whether Landy really is describing a ‘picture theory’ of representation and get straight to the heart of the matter. I think my real worry with Landy’s view is with the idea that the relations among the sensory components of an intuition are conceptual (but apparently, again, not spatial). This concerns me for two reasons. I’ll explain them both, and this will lead me to propose (what I take to be) a friendly modification of Landy’s view.
First, I worry that Landy’s inferentialist interpretation over-intellectualises intuitions. This is, by the way, a common accusation levied against conceptualist interpretations. I don’t know whether all conceptualists are guilty of this, but I think it is going on in Landy’s reading. The problem, as I see it, is that by focusing on conceptual form and ignoring both spatial form as well as the phenomenal availability of sensations, we have an account only of how objects are thought, but not of how they are seen (or otherwise perceived). Although Landy does admit that intuitions have a non-conceptual component—sensations—since sensations are not phenomenally present and are known of instead only through the sense impression inference, it follows that sensations and their qualities aren’t truly a part of the content of conscious experience.
Evidence that Landy is guilty of over-intellectualising intuitions can be found in some of his examples. At one point, he analyses the proposition “Every metal is a body” and claims that “‘every metal’ is an intuition of all metals” (p. 95). This can’t be right though. Kant is adamant that intuitions are singular representations that pick out individual objects. No intuition could ever function to represent all metals. It would seem that Landy is treating possible intuitive representational content as coextensive with possible conceptual representational content. But this is a mistake, since only concepts can represent generally, and the representation of ‘all metals’ is a general representation.
Later, Landy offers an interesting discussion of the different ways that a scientific laboratory would be perceived by an expert and a non-expert. Although overall I found this discussion to be illuminating, I was surprised at Landy’s claims about the intuitions formed by the expert. He says that, when perceiving a cloud chamber with a streak in it, the expert physicist will form the intuition “‘some-sub-atomic-particle’ or ‘that electron’” (p. 148). I have two problems with this example. First, intuitions are supposed to refer to their objects directly and immediately. But the quantifier ‘some’ indicates an indirect reference, perhaps something like the definite description “whatever sub-atomic particle caused this streak”. Although we can make conceptual use of the logical quantifier ‘some’ and of indirect definite descriptions, intuitions (as I understand them) are not supposed to possess that kind of content. Second (and this will perhaps be more controversial), since in this example no electron is actually seen (rather, the effect of the electron’s motion through the cloud chamber is seen), I do not think that there can be an intuition of an electron at all. Rather, the only intuition will be the intuition of the (perceivable) streak through the cloud chamber. This intuition might license an inference to a belief about the presence of an electron, but it will not itself count as an intuition of an electron. In general, I would argue, there cannot be intuitions of things (like electrons) that are not perceivable. But once again, it looks like Landy’s conceptualism leads him to ask too much of intuitions and their representational content.
Second, I have concerns about how to make sense of the claim that intuitions are conceptually/inferentially structured collections of sensations. It sometimes sounds as though Landy is suggesting that there are inferential relations between sensations as the constituents of intuitions, or between intuitions themselves. This would not amount to a workable theory, however, because sensations and intuitions (even conceptually structured intuitions) do not have propositional form. But there can be inferential relations only between representations that do have propositional form. Although the imagination might be able to make associative connections between sensations and intuitions, it will not be possible for the understanding to infer from one sensation to another or from one intuition to another. Inferential moves can be made only from judgement to judgement.
However, closer inspection reveals that this is not quite what Landy means when he describes intuitions as conceptually structured collections of sensations (see especially pp. 132–54). What he really means, I think, is that the process by which the imagination collects and orders sensations together is conceptually guided. This process is ‘blind’ insofar as it takes place automatically, without any explicit rule-following on the part of the imagination—as Landy notes, it’s “rule-governed” without being “rule-following” (see the insightful discussion at pp. 146ff.). Yet there is still a conceptual normativity at work. The result is that the form instantiated by the intuition (i.e. the organised collection of sensations) has a structure that will mirror the structure of judgements that are subsequently formed in response to the intuitions. These judgements are then themselves supposed to mirror the structures of objects themselves in the world. Hence we have a threefold correspondence: between the structure of sensations in intuitions, the structure of concepts in judgements, and the structure of objects in the world.
Putting it this way makes the view much more palatable to me. However, I also think that it indicates that Landy’s characterisation of his view needs to be modified in an important respect. If I understand him correctly, he’s being misleading when he says that intuitions have conceptual structure. Rather, it sounds instead like what he should really be saying is that intuitions have a structure that is not itself inferential, but which results from a conceptually-guided process. In general, to say that something was formed by a conceptually-guided process is not to say that it has conceptual form. To 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. It’s the same for the intuitions resulting from the inferentially-guided, blind operations of the imagination: if sensations are combined into an intuition of a house because the imagination was guided by the concept ‘house’, it need not follow that the sensations themselves come to have conceptual structure.
If I am right that Landy should really be saying this instead (that sensations are organised in accordance with inferential rules, but not that they are themselves inferentially structured), then it raises the question of what form intuitions have instead. I have already indicated my preferred answer to this question: the form of intuition is space (and time, but I’ll focus on space). That is, Landy could have (and I think should have) said that intuitions are spatially organised collections of sensations which have been organised by a conceptually/inferentially-guided synthesis. This would still count as a kind of conceptualist reading insofar as the content of the intuition depends directly on conceptual-guidedness, and hence intuitions would not be mere brute and mechanical deliveries of receptivity. But it would no longer involve the claim that sensations are given a conceptual form.
Here’s an example of how an intuition of a house could be formed on this modified version of Landy’s interpretation (I roughly follow his account of the A-Deduction’s ‘threefold synthesis’ here; see pp. 132–41). When I perceive the house, my eyes quickly scan the various entities in my vicinity (windows, siding, door, car in driveway, etc.). This occurs over time, and since sensations fade the instant I look elsewhere, if I am to end up with a full representation of a house, I must be able to recall the sensations that are not currently present in my visual field. This ‘reproduction’ must be guided by a forward-looking use of the concept that will give the intuition its character. The concept ‘house’ must be at work so that the imagination reproduces the doors and windows, but not the car parked on the driveway or the dog on the lawn. Finally, with all of these various sensations available in the mind (some actually present, others reproduced), they are combined into a spatially organised picture of the house. This intuition can then have concepts applied to it, or judgements made about it. The fact that this conceptualisation is possible at all is due to the fact that the intuition was formed in accordance with the concept ‘house’ (even if the structure of the sensations making up the intuition isn’t a conceptual structure).
I am optimistic that with this interpretation he can still get most of the conclusions he wants. It seems that he can still tell an inferentialist story about the concepts of substance, causality, the external world, and the persistent self even after admitting that the form of intuition is spatiality, not conceptuality. Though I do admit that he has thought about these issues more carefully than I have, and so if there are costs to modifying the interpretation in the way I propose, I look forward to Landy pointing them out in his response.
Before closing, it may be worth returning briefly to Hume. Hume thought that the components of a complex representation needed to have spatial structure in order to represent spatial objects. Now I’ve made the same claim about Kant. Does this mean that (my) Kant would be susceptible to the same objections raised against Hume? I don’t think so. Hume faced the problem of representational overflow (discussed above) because he thought that representational content depends exclusively on the features of the mental picture: all features of the picture are actively representing corresponding features in the object, which makes attention to any one feature (at the exclusion of the rest) impossible. I have not attributed this view to Kant. On my view, even though intuitions are spatial pictures that represent spatial objects, there is still important additional work to be done by concepts, which are distinct from the features of the picture, even though they are partially responsible for the structure of these pictures. Concepts allow for the possibility of focusing attention on specific features of represented objects, and this forestalls the threat of representational overflow.
In conclusion, I will say that despite my disagreements with Landy about some important details in Kant’s theory of intuition, I found the book on the whole to exhibit impressively sustained and careful argumentation. Although I remain sceptical about the extent to which Kant should be read in Sellarsian terms (I am okay with this to some degree), it was really nice to see the case being made for a Sellarsian Kant in such a detailed and thorough manner. Where many appeal to Sellars only to explain some of what’s going on in the Transcendental Aesthetic and the early sections of the Transcendental Logic, Landy carries the interpretation all the way through the Analogies and Paralogisms and even to some issues with transcendental idealism itself. Thus the claim that Kant can be read through a Sellarsian lens needn’t remain a mere promissory note, and Landy demonstrates the real cash value of such an interpretation. He should be lauded for this.
Invited: 21 October 2015; received: 8 November 2016.
Allais, L. (2009), ‘Kant, Non-Conceptual Content, and the Representation of Space’, Journal of the History of Philosophy 47(3): 383–413.
Jankowiak, T. (2014), ‘Sensations as Representations in Kant’, British Journal for the History of Philosophy 22(3): 492–513.
McDowell, J. (1998), ‘Having the World in View: Sellars, Kant, and Intentionality’ [Woodbridge Lectures], Journal of Philosophy 95(9): 431–92.
Pippin, R. (1982), Kant’s Theory of Form: An Essay on the Critique of Pure Reason (New Haven: Yale University Press).
Thompson, M. (1972), ‘Singular Terms and Intuitions in Kant’s Epistemology’, Review of Metaphysics 26(2): 314–43.
Watkins, E. (2012), ‘Kant, Sellars, and the Myth of the Given’, Philosophical Forum 43(3): 512–31.
© Tim Jankowiak, 2017.
Tim Jankowiak is Assistant Professor of Philosophy at Towson University, Baltimore, US. The primary focus of his research is on Kant’s metaphysics, epistemology and philosophy of mind and early modern philosophy. He has published in, among other, Kantian Review, British Journal for the History of Philosophy and Journal of Philosophical Research. His article ‘Kantian Phenomenalism Without Berkeleyan Idealism’ is forthcoming in Kantian Review.