Kant’s Transcendental Idealism

Draft -- commissioned for Stanford Encyclopedia of Philosophy; awaiting final approval by editors

In the Critique of Pure Reason Kant argues that space and time are

merely formal features of how we perceive (intuit) objects, not things

in themselves that exist independently of us, or properties or relations

among them. Objects in space and time are said to be ‘appearances’,

and we know nothing of the things in themselves of which they are

appearances. Kant calls this doctrine (or set of doctrines)

‘transcendental idealism,’ and ever since the publication of the first

edition of the Critique of Pure Reason in 1781, Kant’s readers have

wondered, and debated, what exactly transcendental idealism is, and

have developed quite different interpretations. Some, including many

of Kant’s contemporaries, interpret transcendental idealism as

essentially a form of phenomenalism, similar in some respects to that

of Berkeley, while others think that it is not a metaphysical or

ontological theory at all. There is probably no major interpretive

question in Kant’s philosophy on which there is so little consensus.

This entry provides an introduction to the most important Kantian

texts, as well as the interpretive and philosophical issues surrounding

them.

1

1. Appearances and Things in Themselves

In the first edition of the Critique of Pure Reason, published in 1781,

Kant argues for a surprising set of claims about space, time and

objects:

Space and time are merely the forms of our sensible intuition of objects. They are not beings that exist independently of our intuition (things in themselves), nor are they properties of, nor relations among, such beings. (A26, A33)

The objects we intuit in space and time are appearances, not objects that exist independently of our intuition (things in themselves). This is also true of the mental states we intuit in introspection; in ‘inner sense’ (introspective awareness of my inner states) I intuit only how I appear to myself, not how I am ‘in myself.’ (A37-8, A42)

We can only cognize objects that we can, in principle, intuit. Consequently, we can only cognize objects in space and time, appearances. We cannot cognize things in themselves. (A239)

Nonetheless, we can think about things in themselves using the

categories (A254).

Things in themselves affect us, activating our sensible faculty

(A190, A387).1

In the Fourth Paralogism Kant defines ‘transcendental idealism’:

I understand by the transcendental idealism of all appearances [Erscheinungen] the doctrine that they are all together to be regarded as mere representations and not as things in themselves [nicht als Dinge an sich selbst ansehen], and accordingly that space and time

1 Ever since the publication of the Critique, this claim, the so-called ‘doctrine of noumenal affection, has been especially controversial since it apparently involves predicating a category (cause-effect) of things in themselves. I discuss this controversy below, in section 3.4.

2

are only sensible forms of our intuition, but not determinations given for themselves or conditions of objects as things in themselves [als Dinge an sich selbst]. (A369)2

Ever since 1781, the meaning and significance of Kant’s

‘transcendental idealism’ has been a subject of controversy. Kant’s

doctrines raise numerous interpretive questions, which cluster around

three sets of issues:

(a)The nature of appearances. Are they (as Kant sometimes suggests) identical to representations, i.e. states of our minds? If so, does Kant follow Berkeley in equating bodies (objects in space) with ideas (representations)? If not, what are they, and what relation do they have to our representations of them?

(b)The nature of things in themselves. What can we say positively about them? What does it mean that they are not in space and time? How is this claim compatible with the doctrine that we cannot know anything about them? How is the claim that they affect us compatible with that doctrine? Is Kant committed to the existence of things in themselves, or is the concept of a ‘thing in itself’ merely the concept of way objects might be (for all we know)?

(c) The relation of things in themselves to appearances. Is the appearance/thing in itself an ontological distinction between two different kinds of objects? If not, is it a distinction between two aspects of one and the same kind of object? Or perhaps an adverbial distinction between two different ways of considering the same objects?

Sections 2-6 examine various influential interpretations of

transcendental idealism, focusing on their consequences for (a)-(c).

Section 7 is devoted more narrowly to the nature of things in

themselves, topic (b), and the related Kantian notions: noumenoa, and

2 The Critique is quoted from the Guyer-Wood translation, Kant (1998).

3

the transcendental object. The primary focus will be the Critique of

Pure Reason itself; while transcendental idealism, arguably, plays an

equally crucial role in the other Critiques, discussing them would take

us too far afield into Kant’s ethics, aesthetics, and teleology.3 While

transcendental idealism is a view both about space and time, and thus

of objects of outer sense as well as inner sense (my own mental

states), this entry will focus on Kant’s views about space and outer

objects. Kant’s transcendental idealist theory of time is too intimately

tied up with his theory of the self, and the argument of the

transcendental deduction, to discuss here.4

Before discussing the details of different interpretations, though, it

will be helpful if readers have an overview of some relevant texts and

some sense of their prima facie meaning. These interpretation of

these texts offered in this section is provisional; later, we will see

powerful reasons to question whether these interpretations are

correct. Since some scholars claim there is a change in Kant’s

doctrine from the A edition of 1781 to the B edition of 1787, we will

begin by restricting attention to the A edition. Section 2.4 discusses

what relevance the changes made in the B edition have for the

3 In the Critique of Practical Reason, transcendental idealism is invoked to secure the possibility of the highest good (Ak. 5:134-136); in the Critique of Judgment, it is invoked to prove the possibility of a reflective judgment of taste (Ak. 5:338-351) and the possibility of a reflective judgment of purposiveness (Ak. 5:405-415).

4 See, however, Falkenstein (1991); Van Cleve (1999), 52-61; and Dunlop (2009).

4

interpretation of transcendental idealism. However, following

standard scholarly practice, for passages present in both editions, the

A page number followed by the B page number is given (e.g.

A575/B603). Works other than the Critique are cited by volume in the

‘Academy’ edition of Kant’s work (Ak.), followed by the page number.

At the end of this article can be found a guide to all the editions and

translations of Kant used in its preparation.

1.1 Transcendental Realism and Empirical Idealism

One promising place to begin understanding transcendental idealism

is to look at the other philosophical positions from which Kant

distinguishes it. In the Fourth Paralogism, he distinguishes

transcendental idealism from transcendental realism:

To this [transcendental] idealism is opposed transcendental realism, which regards space and time as something given in themselves (independent of our sensibility). The transcendental realist therefore represents outer appearances (if their reality is conceded) as things in themselves [Dinge an sich selbst], which would exist independently of us and our sensibility and thus would also be outside us according to pure concepts of the understanding. (A369)

Transcendental realism, according to this passage, is the view that

objects in space and time exist independently of our experience of

them, while the transcendental idealist denies this. This point is

reiterated later in the Critique when Kant writes:

We have sufficiently proved in the Transcendental Aesthetic that everything intuited in space or in time, hence all objects of an

5

experience possible for us, are nothing but appearances, i.e. mere representations, which, as they are represented, as extended beings or series of alterations, have outside our thoughts no existence grounded in itself. This doctrine I call transcendental idealism. The realist, in the transcendental signification, makes these modifications of our sensibility into things subsisting in themselves, and hence makes mere representations into things in themselves [Sachen an sich selbst]. (A491/B515)5

Appearances exist at least partly in virtue of our experience of them,

while the existence of things in themselves is not grounded in our

experience at all.6 Kant calls transcendental realism the “common

prejudice” (A740/B768) and describes it as a “common but fallacious

presupposition” (A536/B564).7 Transcendental realism is the

commonsense pre-theoretic view that objects in space and time are

‘things in themselves’, which Kant, of course, denies.

Kant also distinguishes transcendental idealism from another

position he calls ‘empirical idealism’:

One would also do us an injustice if one tried to ascribe to us that long-decried empirical idealism that, while assuming the proper reality of space, denies the existence of extended beings in it, or at least finds this existence doubtful, and so in this respect admits no satisfactory provable distinction between dream and truth. As to the appearances of inner sense in time, it finds no difficulty in them as real things, indeed, it even asserts that this inner experience and it alone gives sufficient proof of the real existence of their object (in itself) along with all this time-determination. (A491/B519)

5 In the B Edition, Kant adds a footnote here, pointing out that ‘formal’ idealism might be a better term for this view, to distinguish it from ‘material’ idealism (which he elsewhere calls ‘empirical’ idealism).

6 Cf. A369, A492/B521, A493/B522.

7 Cf. Allison (2004), 22.

6

Empirical idealism, as Kant here characterizes it, is the view that all

we know immediately (non-inferentially) is the existence of our own

minds and our temporally ordered mental states, while we can only

infer the existence of objects ‘outside’ us in space. Since the

inference from a known effect to an unknown cause is always

uncertain, the empirical idealist concludes we cannot know that

objects exist outside us in space. Kant typically distinguishes two

varieties of empirical idealism: dogmatic idealism, which claims that

objects in space do not exist, and problematic idealism, which claims

that objects in space may exist, but we cannot know this.8 Although

he is never mentioned by name in the A Edition, Berkeley seems to be

Kant’s paradigm dogmatic idealist, while Descartes is named as the

paradigm problematic idealist.9

Transcendental idealism is a form of empirical realism because it

entails that we have immediate (non-inferential) and certain

knowledge of the existence of objects in space merely through self-

consciousness:

[. . .] external objects (bodies) are merely appearances, hence also nothing other than a species of my representations, whose objects are something only through these representations, but are nothing

8 See A377.

9 See especially B274, but also B71. Berkeley is also cited as a dogmatic idealist in the Prolegomena (Ak. 4:293, 374) and Refl. 6311 (Ak. 18:610) and the metaphysics lectures (Ak. 28:680, 29:928). Beiser (2002, 79-80) argues that in the A Edition Kant has Leibniz in mind as the representative dogmatic idealist.

7

separated from them. Thus external things exist as well as my self, and indeed both exist on the immediate testimony of my self-consciousness, only with this difference: the representation of my Self, as the thinking subject is related merely to inner sense, but the representations that designate extended beings are also related to outer sense. I am no more necessitated to draw inferences in respect of the reality of external objects than I am in regard to the reality of my inner sense (my thoughts), for in both cases they are nothing but representations, the immediate perception (consciousness) of which is at the same time a sufficient proof of their reality. (A370-1)

Merely through self-conscious introspection I can know that I have

representations with certain contents and since appearances are

“nothing other than a species of my representations” this constitutes

immediate and certain knowledge of the existence of objects in space.

Understanding transcendental idealism requires understanding the

precise sense in which things in themselves are, and appearances are

not, ‘external to’ or ‘independent’ of the mind and Kant draws a

helpful distinction between two senses in which objects can be

‘outside me’:

But since the expression outside us carries with it an unavoidable ambiguity, since it sometimes signifies something that, as a thing in itself [Ding an sich selbst], exists distinct from us and sometimes merely that belongs to outer appearance, then in order to escape uncertainty and use this concept in the latter significance — in which it is taken in the proper psychological question about the reality of our outer intuition — we will distinguish empirically external objects from those that might be called ‘external’ in the transcendental sense, by directly calling them ‘things that are to be encountered in space.’ (A373)

In the transcendental sense, an object is ‘outside me’ when its

existence does not depend (even partly) on my representations of it.

The empirical sense of ‘outside me’ depends upon the distinction

8

between outer and inner sense. Inner sense is the sensible intuition

of my inner states (which are themselves appearances); time is the

form of inner sense, meaning that all the states we intuit in inner

sense are temporally ordered. Outer sense is the sensible intuition of

objects that are not my inner states; space is the form of outer sense.

In the empirical sense, ‘outer’ simply refers to objects of outer sense,

objects in space. Transcendental idealism is the view that objects in

space are ‘outer’ in the empirical sense but not in the transcendental

sense. Things in themselves are transcendentally ‘outer’ but

appearances are not.

1.2 The Empirical Thing in Itself

Just as Kant distinguishes a transcendental from an empirical sense of

‘outer’ he also distinguishes a transcendental version of the

appearance/thing in itself distinction (the distinction we have been

concerned with up to now) from an empirical version of that

distinction:

We ordinarily distinguish quite well between that which is essentially attached to the intuition of appearances, and is valid for every human sense in general, and that which pertains to them only contingently because it is not valid for the relation to sensibility in general but only for a particular situation or organization of this or that sense. And thus one calls the first cognition one that represents the object in itself, but the second one only its appearance. This distinction, however, is only empirical. If one stands by it (as commonly happens) and does not regard that empirical intuition as in turn mere appearance (as ought to happen) , so that there is nothing to be

9

encountered in it that pertains to any thing in itself, then our transcendental distinction is lost, and we believe ourselves to cognize things in themselves, although we have nothing to do with anything except appearance anywhere (in the world of sense), even in the deepest research into its objects. Thus, we would certainly call a rainbow a mere appearance in a sun-shower, but would call this rain the thing in itself, and this is correct, as long we understand the latter concept is a merely physical sense, as that which in universal experience and all different positions relative to the senses is always determined thus and not otherwise in intuition. But if we consider this empirical object in general and, without turning to its agreement with every human sense, ask whether it (not the raindrops, since these, as appearances, are already empirical objects) represents an object in itself, then the question of the relation of the representation to the object is transcendental, and not only these drops are mere appearances, but even their round form, indeed even the space through which they fall are nothing in themselves, but only mere modification of our sensibility or foundations of our sensible intuition; the transcendental object, however, remains unknown to us. (A45-46/B62-63)10

In the empirical case, the distinction seems to be between the

physical properties of an object and the sensory qualities it presents

to differently situated human observers. This requires distinguishing

between what is “valid for every human sense in general” and what

“pertains to [objects] only contingently because [of] . . . a particular

situation or organization of this or that sense” (A45/B62). The

distinction seems to be that some properties of objects are

represented in experience just in virtue of the a priori forms of

experience, and thus have inter-subjective validity for all cognitive

subjects, while some properties depend upon the particular

constitution of our sense-organs.11 The ‘empirical thing in itself’ is the

10 Cf. the discussion of the rose at A29-30/B45, as well as A257/B313.

11 Cf. A226/B273.

10

empirical object qua bearer of the former set of properties, while the

‘empirical appearance’ is the empirical object qua bearer of all of its

properties, including the latter. For instance, the empirical ‘rainbow

in itself’ is a collection of water droplets with particular sizes and

shapes and spatial relations, while the empirical ‘rainbow appearance’

is the colorful band we see in the sky.12

For our purposes, the importance of this distinction is two-fold.

Firstly, the (transcendental) distinction is not the ordinary distinction

between how objects appear to us in sense perception and the

properties they actually have. Kantian appearances are not the

objects of ordinary sense perception, for Kant holds that appearances

in themselves (things in themselves, in the empirical sense) lack

sensory qualities like color, taste, texture, etc. In scientific research,

we may discover how appearances are in themselves (in the empirical

sense) but in so doing all we discover is more appearance (in the

transcendental sense); scientific investigation into the ultimate

constituents or causal determinants of objects only reveals more

appearance, not things in themselves. Secondly, there is an

appearance/reality distinction at the level of appearances. This

12 The empirical thing in itself corresponds roughly to Lockean primary qualities, while the empirical appearance corresponds roughly to its secondary qualities. For Kant’s own comparison of his idealism to that Lockean distinction see Prolegomena (Ak. 4:289); Allais (2007) is a sophisticated discussion of Kant’s secondary quality analogy.

11

provides a further sense in which Kant is an ‘empirical realist’:

appearances in themselves have properties quite different than they

seem to have in sense perception.

Kant’s empirical realism – not in his technical sense, but in the

broader sense that he accepts an appearance/reality distinction at the

level of appearances13 – is further deepened by his scientific realism:

he accepts the existence of unobservable entities posited by our best

scientific theories and holds that these entities are appearances

(because they are in space).14 Earlier, we saw texts whose prima facie

meaning is that appearances exist, at least partly, in virtue of the

contents of our representations of them. But it is clear that Kant

cannot hold that the existence of an object in space is grounded in our

direct perception of that object, for that would be incompatible with

the existence of unperceived spatial objects.

2. The Feder-Garve Review and Kant’s Replies

13 See Abela (2002).

14 E.g. magnetic matter (A226/B273) and ‘lamellae’ (Discovery, Ak. 8:205), light particles posited by Newton. See Langton (1999), 186-204.

12

The first published review of the Critique of Pure Reason, by Feder

and Garve, accuses Kant of holding a basically Berkeleyan

phenomenalist conception of objects in space. Feder and Garve were

not the only ones to read Kant as a phenomenalist. The

phenomenalist reading was so widespread and influential that it

became the default interpretation for generations after the

publication of the Critique. In fact, many of the key figures in German

philosophy in 1781 and after (e.g. Mendelssohn, Eberhard, Hamann,

Jacobi, Fichte, Schelling, and Hegel) take the phenomenalist or

‘subjectivist’ reading of Kant for granted and think this is precisely

why Kant must be ‘overcome.’ The assumption that Kant is a

subjectivist about appearances is a major impetus in the development

of German idealism. 15

However, the phenomenalist reading of transcendental idealism

has been challenged on many fronts, both as an interpretation of Kant

and (often on the assumption that it is Kant’s view) on its own

philosophical merits. In this section I explain the origin of the

phenomenalist reading in the Feder-Garve review and its basis in the

text of the Critique. In the next section I argue that the

phenomenalist reading is much more defensible as an interpretation

of Kant than is often appreciated. In section 3.4 I explore influential

15 Beiser (1987) and (2002) are particularly illuminating on the influence of the subjectivist reading on the development of post-Kantian German philosophy. See also Vaihinger (1892), vol. 2, 35-55.

13

objections by Kant’s contemporaries to transcendental idealism, on

the assumption that the phenomenalist interpretation of that doctrine

is correct, which were later taken up as criticisms of the

phenomenalist interpretation itself. In section five I introduce a

theme that I explore in greater detail in later sections: the

development of non-phenomenalist interpretations of Kant’s

transcendental idealism.

2.1 The Feder-Garve review

Although it is uncharitable and, on some points, simply mistaken, the

first published review of the Critique, originally written by Christian

Garve and then substantially revised, and shortened, by J.G.H. Feder,

raised an issue that has been discussed ever since. The Göttingen or

Feder-Garve review, as it is now known, claims that Kantian

‘transcendental’ idealism is just idealism of a familiar Berkeleyan or

phenomenalist variety:

According to the author, experience, contrary to mere fancy and dreams, is [composed] of sensible intuitions combined with concepts of understanding. We admit, however, that we do not comprehend how the distinction between what is actual from what is imagined and merely possible, a distinction that is generally so easy for human understanding, could be sufficiently grounded in the mere application of concepts of understanding without assuming one mark of actuality in sensation itself. This is the case particularly in view of the fact that for those who are dreaming as well as those who are awake, visions and fantasies can occur as outer appearances in space and time, and, in general, as combined with one another in a most orderly fashion, sometimes even to all appearances in a more orderly fashion than

14

actual events [. . .] when, to assume the most extreme position with the idealist, everything of which we can know and say something is merely representation and law of thought, when representations in us, modified and ordered in accord with certain laws are just that which we call object and world, why then the fight against this commonly accepted language, why then and from where this idealist differentiation?16

Feder-Garve read Kant as identifying objects in space with our outer

representations, i.e. our perceptions ‘of’ such objects. They are aware

that Kant does not think just any representation is an object; objects

are given only in experience. The subject’s mind generates experience

by synthesizing sensory contents, according to the a priori forms of

space and time and the categories, into object-experiences. Feder-

Garve’s point is that dreams and hallucinations can be just as

internally unified and coherent as veridical experience (think of the

Matrix). Dreams are spatiotemporal, and can have the unity

prescribed by the categories. If objects are just collections of unified

experiences, then there is no difference in principle between veridical

experience and a highly coherent dream. Objects are just collections

of highly internally unified representations. At one point, they note in

passing that “one basic pillar of the Kantian system rests on these

concepts of sensations as mere modifications of ourselves (on which

16 Sassen 2000, 54-58The same volume also contains a translation of Garve’s much longer original draft; the originals can be found in Karl Vorländer’s edition of the Prolegomena (Hamburg: Meiner, 1969), 167-74. In a letter to Kant, Garve claims that the published review is mainly Feder’s fault (Ak. 10: 331). However, Beiser (2002, 88) points out that nearly two thirds of the published review comes from the Garve’s original text.

15

Berkeley, too, principally builds his idealism).”17 The implication is

clear: Kant has a basically Berkeleyan conception of objects in space.

First of all, it should be noted that the Feder-Garve view, while not

exactly an exercise in interpretive charity, is not without basis in

claiming that there is a deep similarity between Berkeley and the

Critique.18 First of all, Kant repeatedly claims that empirical objects

are representations. For instance, in the Transcendental Aesthetic he

writes that “what we call outer objects are nothing other than mere

representations of our sensibility” (A30/B45) and in the Fourth

Paralogism he writes: “external objects (bodies) are merely

appearances, hence also nothing other than a species of my

representations” (A370).19 Since ‘representation’ [Vorstellung] is

Kant’s term for what Berkeley calls ‘ideas’, this seems at least

perilously close to the Berkeleyan view that bodies are collections of

ideas. Secondly, the A Edition is full of passages that can easily

suggest a phenomenalist view of objects in space, such as:

Why do we have need of a doctrine of the soul grounded merely on pure rational principles? Without doubt chiefly with the intent of securing our thinking Self from the danger of materialism. But this is achieved by the rational concept of our thinking Self that we have given. For according to it, so little fear remains that if one took matter away then all thinking and even the existence of thinking beings would be abolished, that it rather shows clearly that if I were to take

17 Sassen 2000, 53.

18 This point is brought out well in Beiser (2002), 49-52.

19 See also A30/B45, A104 and A375n, A490, A498, A563.

16

away the thinking subject, the whole corporeal world would have to disappear, as this is nothing but the appearance in the sensibility of our subject and one mode of its representations. (A383)

One must note well this paradoxical but correct proposition, that nothing is in space except what is represented in it. For space itself is nothing other than representation; consequently, what is in it must be contained in representation, and nothing at all is in space except as it is really represented in it. A proposition which must of course sound peculiar is that a thing exists only in the representation of it; but it loses its offensive character here, because the things with which we have to do are not things in themselves but only appearances, i.e., representations. (A374n) 20

On one plausible reading of these passages, Kant is claiming that all

there is for objects in space to exist is for us to have experiences as of

objects in space. Consequently, if we did not exist, or did not have

such experiences, these objects would not exist. The Feder-Garve

interpretation of transcendental idealism is not without merit.21

2.2 Varieties of Phenomenalism

Phenomenalism can mean many things, and later I will explore these

meanings in detail, but for now it is worth distinguishing at least three

different things we might mean by phenomenalism:

20 Cf. A490-1/B518-9, A520/B492-A521/B493, A494/B522.

21 As Beiser (2002) points out, Jacobi, whose influential objections to Kant’s idealism are discussed in section 3.4, cites the following passages in defense of the phenomenalist interpretation: A370, A372-3, A374-75n, A378, A379-80, A36-37, A37n, A491, A101, A125, A126-7.

17

(1) Objects in space are identical to (unified collections of) our representations of them.22

(2) Objects in space exist solely in virtue of the contents of our representations of them. They possess all of their properties solely in virtue of the contents of those representations.

(3) Objects in space exist partly in virtue of the contents of our representations of them. They possess their ‘core physical properties’ solely in virtue of the contents of those representations.

By ‘core physical properties’ I mean the properties that appearance

have ‘in themselves’ according to Kant, roughly, Lockean primary

qualities.23 Feder-Garve accuse Kant of holding (1), which I will call

‘identity phenomenalism.’ But even if he did not hold that extreme

view, he might hold one of the weaker views listed here. Claim (2) is a

quite strong form of phenomenalism, for it entails that, in some sense,

all there is to objects is our representations of them, although they

are not literally identical to those representation. I will call this

‘strong phenomenalism.’ The exact meaning of Berkeley’s own views

about bodies is unclear, and not the subject of this entry. But it is not

22 Discussions of Kant’s alleged phenomenalism have been somewhat distorted in Anglophone literature by the emphasis on the semantic version of phenomenalism made popular by (among others) Russell and Carnap according to which sentences about objects are equivalent in meaning to complex sentences about mental states. It is implausible that Kant intended such a semantic reduction. Cf. Bennett (1971), 136-7 and Allison (2004), 38. Van Cleve (1999), 8-12 requires more minimally that facts about objects be ‘derivable’ from facts about mental states, but if ‘derivable’ here means something like ‘derivable using logic and definitions’ it is still too strong for Kant.

23 See Locke, Essay concerning Human Understanding, book II, chapter VIII.

18

implausible to read Berkeley as holding (2). However, claim (3), while

very controversial and (arguably) extremely counter-intuitive, is

weaker. It allows that there may be more to the existence of an object

in space than our representing them, and it allows that there may be

aspects or properties of objects that they possess independently of

how we represent them. I will call it ‘qualified phenomenalism.’ In

discussing the debate about Kant’s alleged phenomenalism, and

Kant’s own responses to the Feder-Garve review, it will help to have

these distinctions in mind.

2.3

Kant Strikes Back

Kant’s was apoplectic that Feder and Garve had, apparently, not made

any serious attempt to even understand the Critique, or to present its

contents accurately to their readers. He penned a response to the

review, published as an Appendix to the Prolegomena. In the

Appendix, and in the text of the Prolegomena itself, Kant explains

what he sees as clear differences between his own view and

Berkeley’s. First, Kant identifies idealism as the doctrine that “all

cognition through the senses and experience is nothing but sheer

illusion, and there is truth only in the ideas of pure understanding and

reason” (Ak. 4:374) and points out that, in this sense, his view is not

19

idealism at all because the Critique consistently maintains that bodies

exist in space and that we have immediate (non-inferential)

knowledge of them. Secondly, Kant points out that, in contrast to

Berkeley’s empiricist theory of ideas, space and time constitute

necessary and a priori forms for any experience of objects (for minds

like ours). Consequently, agreement with the structural forms of

space and time provides a necessary and a priori cognizable set of

rules for distinguishing between truth from illusion in experience.24

Thirdly, Kant points out that his idealism is merely formal: he has

argued only that the form of objects is due to our minds, not their

matter.25 While the form-matter distinction in Kant’s philosophy is a

complex matter in its own right, I take his point to be that the matter

of experience, the sensory content that is perceptually and

conceptually structured by space and time, and the categories,

respectively, is not generated by the mind itself, but produced in our

minds through affection by mind-independent objects, things in

themselves. As he would write several years later in response to

Eberhard, the Critique “posits this ground of the matter of sensory

representations not once again in things, as objects of the senses, but

in something super-sensible, which grounds the latter, and of which

we can have no cognition” (Discovery, Ak. 8:205). Thus, Kant can

24 Ak. 4:375.

25 Cf. Kant’s Dec. 4 1792 letter to J.S. Beck (Ak. 11:395).

20

claim that only the form of experience is mind-dependent, not its

matter; the matter of experience depends upon a source outside of the

mind.26

However, Kant’s attempts to distance himself from Berkeley do not

cut nearly as deep as he seems to of think. Regarding the first point,

Kant’s definition of idealism in the Appendix (quoted above) does not

apply to Berkeley, nor does the definition in the body of the

Prolegomena: “the claim that there are none other than thinking

beings; the other things that we believe we perceive in intuition are

only representations in thinking beings, to which in fact no object

existing outside these beings corresponds” (Ak. 4:289). One of the

main points of Berkeley’s philosophical project is to defend the

existence of bodies in space, while denying the philosophical

misinterpretation of what this existence amounts to: the existence of

non-thinking substances. Berkeley does not deny that bodies exist; he

claims that bodies cannot exist without minds to perceive them,

something that Kant himself also seems to accept (see the texts

quoted in the previous section). In fact, he constantly contends that

his theory is the only way to avoid what Kant calls ‘problematic’

26 The relation of Kant to Berkeley has been extensively discussed in the secondary literature. Scholars who agree with Kant that his idealism is fundamentally different than Berkeley’s include Wilson (1971), Allison (1973), Walker (1985), Beiser (2002, 82-103), and Emundts (2008). The similarity of Kant to Berkeley is argued for by Turbayne (1955) and (1969).

21

idealism: we do not know whether bodies exist.27 That Kant would

describe Berkeley as an idealist in this sense (what he elsewhere

designates a ‘dogmatic idealist’) raises the suspicion that he had not

read Berkeley’s main philosophical writings.28

If this is correct, then Kant’s insistence that he does not share a

Berkeleyan view about the ontological status of bodies is simply not

probative as to whether transcendental idealism is a kind of

phenomenalism. Since the misinterpretation of Berkeley as holding

that sense perception is illusory and that bodies do not exist was

widespread in Germany in the eighteenth-century,29 it is quite possible

that Kant shares this misinterpretation. It may be that Kant is more

similar to Berkeley than he realizes because he is not familiar with

Berkeley’s actual theory.

Now, in the definition of idealism, by “objects existing outside” our

minds Kant might mean two things. In the terminology he uses in the

A Edition Fourth Paralogism he might mean objects “empirically

27 Berkeley, Treatise on the Principles of Human Knowledge, Part I, § 1, 6, 18, 20, 22-24, 34-38.

28 Beiser (2002, 100-102) convincingly argues that Kant’s interpretation of Berkeley fits his later work Siris rather well (the only Berkeleyan work Kant cites by name) rather than his earlier Principles and Dialogues, on which our modern reading of Berkeley is largely based. But this does not affect my point in the body of the text: if Kant is ignorant of, or simply not talking about, Berkeley’s Principles and Dialogues view, his horror at being identified with the good Bishop is not evidence against the phenomenalist reading.

29 Again, see Beiser (2002).

22

external” to our minds—objects that are spatially distinct from us and

thus ‘outer’—or he might means objects “transcendentally external”

to our minds—objects that do not depend upon our minds at all, things

in themselves (A373). In the previous paragraph, I assumed he meant

that idealists deny the existence of empirically external objects.

(After all, in the Appendix, he defines idealism as the claim that sense

perception is illusory). But Kant might mean that idealists deny the

existence of transcendentally external objects, things in themselves,

and in this sense Berkeley is, and he is not, an idealist. Kant may be

right to point out that the thing in itself constitutes a clear difference

between his view and Berkeley’s. But the thing in itself does little or

nothing to distinguish Berkeley from Kant on the very issue Feder-

Garve raised: the ontological status of objects in space. The question

is whether Kant is a phenomenalist (of some stripe) about object in

space, not about things in themselves.30

Kant’s second point—that the a priori forms of space and time are

a source of objectivity not available to Berkeley —is largely irrelevant

30 I find it independently implausible that at 4:289 by “objects outside us” Kant means things in themselves. Kant standardly distinguishes between two forms of idealism: dogmatic idealism (which denies the existence of outer objects) and problematic idealism (which is agnostic about the existence of outer objects). Immediately after the definition of idealism at 4:289 Kant reminds us that we cannot know things in themselves. So if he mean “objects outside us” to refer to things in themselves, he would just have committed himself to problematic idealism, but Kant consistently argues that transcendental idealism is the only way to avoid problematic idealism. The issue of idealism, for Kant, is an issue about objects in space.

23

to determining whether Kant is a phenomenalist.31 The Feder-Garve

charge is that Kantian bodies are simply collections of

representations. That space and time are necessary conditions on

experiencing objects entails (at most) that any possible empirical

object is in space and time, and that we have knowledge a priori of

the possible spatiotemporal properties of objects. It does not entail

that those objects are anything more than collections of

representations (including, of course, a priori spatiotemporal

representations). Nor does it entail that there is anything more to the

existence of a body than subjects’ having experiences with a certain

content (including spatiotemporal content). Claiming that experience

has a necessary a priori spatiotemporal form does nothing to undercut

the phenomenalist implications of claiming that bodies are identical to

collections of representations or that they exist in virtue of the

contents of those representations, if, as Feder-Garve allege, Kant is

committed to one of those views.32

31 This point is, I think, missed by the discussion of B70-1 in Allison (2004), 25.

32 Beiser (2002, 98-99) anticipates this objection and replies, on Kant’s behalf, that even if Berkeley intended to establish the reality of objects in space, Kant could plausibly argue that he is committed to their being mere illusions. But Kant’s claim is that Berkeley’s idealism is the denial that objects exist in space. It is not germane for Kant to argue that Berkeley’s idealism entails, given Kantian premises about what constitutes the difference between truth and illusion, that objects are illusory.

24

Nor does Kant’s third point—that Kant’s idealism concerns merely

the form and not the matter of experience—constitute a clear

difference from Berkeley. Berkeley does not claim that human spirits

are the causes of their own ideas; he claims that God acts on human

spirits, causing us to perceive an internally and inter-subjectively

consistent world of ideas. Since Kant’s official doctrine in the Critique

seems to require agnosticism about the ultimate nature of the things

in themselves that causally affect us in experience, it is compatible

with what he says that the noumenal cause of experience is God

himself!

Kant’s argument might be that the matter of experience (its

sensory content) depends upon how our sensibility is affected by

mind-independent objects, things in themselves, while the form of

experienced by our minds alone. Consequently, experience itself

requires the existence of objects ‘outside’ (in the transcendental

sense) the mind. But this would show, at most, that Kant is not a

strong phenomenalist. It does nothing to undercut the interpretation

of him as a qualified phenomenalist. Nor, as I argued in the previous

paragraph, does it succeed in clearly differentiating him from

Berkeley.

Kant’s attempts to distinguish his own transcendental idealism

from a more familiar Berkeleyan conception of the ontology of bodies

are not successful. Nor, I argued, are they incompatible with reading

25

him as a qualified phenomenalist. This, of course, does not settle the

issue; it may be that Kantian appearances are quite different than

bodies, as Berkeley, or even the qualified phenomenalist, conceive

them.33

2.4 Changes in the B Edition

Kant extensively revised certain sections of the Critique for the

second edition (B), published in 1787. It is widely accepted that a

main consideration in these revisions was to avoid the

misunderstanding of his view that had led to the Feder-Garve review.

However, some scholars think that, on this point, there is a difference

in doctrine between the A and B editions: made aware of the

problematic Berkeleyan consequences of the first edition, Kant

endeavored to develop a more realistic view in the B Edition.34 Other

scholars think the difference is largely a matter of presentation: in the

B edition, Kant highlights the more realistic aspects of his view and

downplays its phenomenalistic sides, but the view is basically the

same.35 In the rest of this section, I go through the main textual

33 For important discussions of transcendental idealism in the Prolegomena see Ak. 4:

34 E.g. Guyer (1987).

35 E.g. Allison (2004).

26

changes from 1781 to 1787 and consider what implications they have

for the interpretation of Kant’s idealism. Since Kant made no

significant changes past the Paralogisms chapter, I will not cite

sections that did not undergo substantial revision as evidence; it may

be that Kant would have significantly changed those sections if had

gotten there.36

Kant did, however, make one relatively minor alteration in the

Antinomies chapter that is relevant to our discussion. In the wake of

the Feder-Garve view, Kant evidently felt that ‘transcendental’

idealism may have been a poor choice of name. 37 In the B Edition

Kant adds a footnote to his definition of transcendental idealism at

A491/519 to remark that perhaps he should have called his position

‘critical idealism.’ But that definition of transcendental idealism, as

we saw earlier, was one of the motivations for the phenomenalist

reading in the first place. Kant would have been aware that his early

readers picked up on the phenomenalist implications of this passage,

for Jacobi cites it as evidence that Kant is a phenomenalist! If Kant

wants to prevent the phenomenalist (mis)interpretation popularized

36 On the general topic of the changes from the A to the B edition, see Erdmann (1878).37 Kant also refers to his own position as ‘formal’ idealism in the Prolegomena (Ak. 4:337). Elsewhere, he refers to it as ‘critical’ idealism (Prolegomena 4:294, 375; On a discovery 8:210).

27

by Feder and Garve, why does he add a footnote to, but not otherwise

amend, a definition with apparently phenomenalist implications?

The section which Kant most heavily revised for the B Edition is the

Transcendetnal Deduction, but I do not have space here to discuss the

complex argument of that section, or the differences between the A

Deduction and the B Deduction

2.4.1 Objects as representations

As mentioned earlier, one of the main sources, both in the eighteenth

century and today, for the phenomenalist reading of Kant is Kant’s

tendency to identify empirical objects with representations. But Kant

continues to do this in the B Edition, not only in sections that were

heavily revised for the B Edition38 but even in occurs in passages that

were added to the B Edition.39 This is evidence that there is no

change on the ‘phenomenalism’ issue from A to B.

2.4.2 Preface

38 In the Aesthetic: “what we call outer objects are nothing other than mere representations of our sensibility” A30/B45). See also A189/B235 and A191/B236. 39 E.g. B164

28

The B Preface contains several passages, which some scholars take to

be inconsistent with the phenomenalist reading. I discuss them below

in section 4.1.

2.4.3 Transcendental Aesthetic

The main additional to the B Transcendental Aesthetic is several

pages (B66-69) at the end of the section, which includes this

discussion, a clear reference to the Feder-Garve review:

If I say: in space and time intuition represents both outer objects as well as the self-intuition of the mind as each affects our senses, i.e. as it appears, that is not to say that these objects would e a mere illusion. [. . .] Thus I do not say that objects merely seem to exist outside me or that my soul only seems to be given if I assert that the quality of space and time ≥ in accordance with which, as condition of their existence, I posit both of these — lies in my kind of intuition and not in these objects in themselves. It would be my own fault if I made that which I should count as appearance into mere illusion. (B70-1)

However, this only reiterates themes found in the A edition and in the

Prolegomena. Phenomenalism does not entail that objects in space

are illusions. Later in the paragraph Kant argues that, if we assume

that if space and time exist they must be “infinite substances,” then

we cannot blame Berkeley for concluding that space, time and bodies

are mere illusions; empirical idealism is the right conclusion to draw

from transcendental realism. While Kant is correct in representing

Berkeley later in this paragraph as reacting against the Newtonian

view of space and time as “absolute” entities, he is wrong to

29

characterize Berkeley as concluding that bodies are mere illusions, so

Kant’s dissatisfaction with Berkeley’s own view is not evidence that he

does not have a phenomenalist view of objects in space. The B

Transcendental Aesthetic adds no new evidence against the

phenomenalist reading.

2.4.4 Paralogisms

One main source of the phenomenalist reading is the A Edition Fourth

Paralogism, in which Kant refutes that the Cartesian view that our

inner states are immediately known while the existence of outer

objects can only be known mediately by inference from our inner

states. The Paralogisms section was entirely re-written in the B

Edition, and none of the four B Paralogisms correspond precisely the

fourth A Paralogism. However, a version of the A Paralogism

argument that self-consciousness requires knowledge of objects in

space, reappears as the Refutation of Idealism in the B Edition, to

which I now turn.

2.4.5 Refutation of Idealism

Given its brevity, the Refutation of Idealism, added to the ‘Postulates

of empirical thinking in general’ in the B Edition, is, line for line, one

30

of the most thoroughly commented upon sections of the entire

Critique.40 Kant’s argument, very briefly, is that the existence of

objects in space outside me (‘empirically external’ objects) is a

condition on the possibility of my being conscious of the determinate

temporal relations of my inner states. Consequently, it is impossible

to be a self-conscious subject without there existing objects in space

outside of me, and in being conscious of the temporal relations of my

inner states I am immediately conscious of the existence of these

objects.41 The problem of ‘problematic idealism’— how can I infer the

existence of objects outside of me on the basis of my immediate

knowledge of my inner states? — is based on a false premise.

Nothing about this conclusion, or how Kant argues for it, is prima

facie incompatible with a qualified phenomenalist reading of

transcendental idealism, or even a strong phenomenalist one.42 It is

40 This is appropriate, since Kant seems to have placed a great deal of importance on the Refutation, judging by the number of Reflections about it: 5653-4, 6311-12, 6313-16, 5709, 6317, 6319 and 6323. For critical discussion, see Guyer (1983), (1987), 279-332; Vogel (1993); Hanna (2000); Dicker (2008); Chignell (2010) and Dicker’s reply, (2011).

41 I am thus departing from the interpretation of Guyer (1983) and (1987), according to which Kant’s ultimate intention in the Refutation, only fulfilled in later Reflections, is to prove the existence of objects ontologically independent of the self (‘things in themselves’).

42 There is, however, a lengthy scholarly controversy over whether the Refutation of Idealism is compatible with the A Edition; see Vaihinger (1884) for a survey of the earlier literature and Guyer (1987) for a more recent argument that, with the Refutation, Kant fundamentally alters his earlier position.

31

presumably incompatible with ‘identity’ phenomenalism, since Kant

argues that self-consciousness requires the existence of permanent

objects in space, yet there is no permanent representation in the

mind.43 If objects just are representations, it follows that none of

them are permanent.44 At B274 Kant makes it clear that the ‘idealism’

that intends to refute is idealism as he defined it in the Prolegomena

and the Fourth A Paralogism: the claim that objects in space do not

exist (dogmatic idealism) or at least that we do not know whether they

exist (problematic idealism). The sense of idealism that is at issue in

the phenomenalist reading—empirical objects exist, and exist in virtue

of the contents of experience—is not, apparently, addressed here. On

an extreme phenomenalist reading, all there is to the existence of

empirical objects in space is our having appropriately unified

experiences of them. The phenomenalist can interpret Kant’s

argument in the ‘Refutation’ as an argument that consciousness of the

temporal relations of my inner states requires that these inner states

constitute appropriately unified experiences. Consequently, self-

43 B278

44 See, though, Kant’s long comment on the Refutation in the B Preface (Bxxxix-Bxli), where he remarks: “the representation of something persisting in experience is not the same as the persisting representation.”(Bxli) So there may be room for an identity phenomenalist reading of the Refutation after all.

32

consciousness requires the existence of objects in space (spatially)

outside me.45

In fact, a broadly phenomenalist analysis of objects is even

compatible with Kant’s final note to the argument:

From the fact that the existence of outer objects is required for the possibility of a determinate consciousness of our self it does not follow that every intuitive representation of outer things includes at the same time their existence, for that may well be the mere effect of the imagination (in dreams as well as delusions); but this is possible merely through the preproduction of previous outer perceptions, which, has been shown, are possible only through the actuality of outer objects. Here it had to be proved that inner experience in general is possible only through outer experience in general. Whether this or that putative experience is not mere imagination must be ascertained according to its particular determinations and through its coherence with the criteria of all actual experience. (B278-9)

Kant points to the distinction between ‘mere imagination’ of objects

(dreams, hallucinations, etc.) and veridical perception and claims that

his argument has established that self-consciousness requires the

latter, although self-consciousness only requires veridical perception

‘in general.’ I take this to mean that a self-conscious subject can

sometimes simply be dreaming, but it is impossible that a self-

conscious subject only ever dreams or hallucinates.46 This is

45 I am only claiming that phenomenalism is prima facie compatible with the Refutation; on this point, I am in agreement with Guyer (1983), 337-338. Whether or not the argument of the Refutation must ultimately be interpreted in a way that demands realism (as Guyer argues), I will not attempt to address here.

46 More minimally, Kant might be read as claiming that the idea of a self-conscious subject requires the idea of experience, perhaps only regulatively, but that it is possible for there to be a self-conscious subject who never attains to experience. I worry, though, that this more minimal claim is incompatible with the results of the

33

compatible with a phenomenalist reading because (as discussed in

greater detail in section 3) the phenomenalist can distinguish between

veridical perception and dreaming. The difference is that the

phenomenalist must analyze veridical perception in terms of the

internal unity and coherence among experiences, rather than their

matching some mind-independent object. For instance, in line with

the interpretation sketched above, the phenomenalist can read Kant

as claiming that self-consciousness requires that enough of one’s

experiences cohere to constitute a unified experience.

Kant concludes the passage by saying that whether a perception is

veridical or not must be judged according to “through its coherence

with the criteria of all actual experience.” What this means, and how

a phenomenalist interpretation can make sense of it, I will discuss

below in section 3.

2.4.6 General note to the principles

In the B Edition Kant added a “General Note” to the Principles of

Experience, which some have read as ruling out the phenomenalist

reading, especially the long passage from B291 to 294, from which I

quote an excerpt:

It is even more remarkable, however, that in order to understand the possibility of things in themselves with the categories, and thus

Transcendental Deduction.

34

establish the objective reality of the latter, we do not merely need intuitions, but always outer intuitions (B291). [. . .] This entire remark is of great importance, not only in order to confirm our preceding refutation of idealism, but, even more, when we come to talk of self-cognition form mere inner consciousness and the determination of our nature without the assistance of outer empirical intuition, to indicate to us the limits of the possibility of such a cognition. (B293-4)

Once again, this is a case of Kant emphasizing that his view is not

idealist in the specific sense of idealism we have seen so far—denying

either that objects exist in space or that we can know that they do.

His point is that even understanding our most basic a priori concepts,

the categories, requires applying them to experienced outer objects in

space. The remark about “self-cognition” at the end is a reminder

that inner awareness is dependent upon outer experience; it does not

address whether empirical objects exist in virtue of the contents of

experience.

2.4.7 Phenomena and noumena

Kant extensively revised the section entitled ‘On the grounds of the

distinction of all objects into phenomena and noumena’ in the B

Edition. However, since that section concerns the Kantian notion of a

‘noumenon’ I will reserve discussion of it until section 8, which is

devoted to that notion, its relation to the ‘thing in itself,’ and the

‘transcendental object.’

35

3. Kant as Phenomenalist

So far, we have seen the prima face evidence for the phenomenalist

interpretation of Kant, made famous by Feder-Garve, and Kant’s own

attempts to distance himself from their accusations. However, we

also distinguished three different kinds of phenomenalism: identity

phenomenalism, strong phenomenalism, and qualified

phenomenalism. I argued that Kant’s objections to the Feder-Garve

interpretation address the distinct issue of whether he is an ‘empirical

idealist’ (one who denies or finds doubtful the existence of objects in

space, which he obviously does not) and do not settle the question of

whether he is a phenomenalist. In this section I am going to explore

the interpretation of Kant as qualified phenomenalist, and argue that

this interpretation can answer many of the standard objections to the

phenomenalist reading.

3.1 Appearances = representations ?

While the identity phenomenalist interpretation has found few

defenders among contemporary readers47, it is worth asking why

exactly we should reject the prima facie meaning of the numerous

47 Guyer (1987), 333-336 is a notable exception. For critical discussion, see Allison (2004), 8-9.

36

passages in which Kant equates appearances with representations.

One argument would be that objects of outer sense are in space, so if

objects just are representations then some of our representations are

in space (the representations of outer sense). For instance, the table

in front of me is a representation in my mind and is 2 feet distant from

me. This is sometimes presented as a devastating objection to the

‘identity phenomenalist’ reading but it is unclear why it should be, for

it depends upon the assumption that mental items like

representations cannot be spatial, or stand in spatial relations to one

another. But it is unclear why Kant must accept this. After all, if the

identity phenomenalist reading is correct, ordinary objects just are

states of our minds, so many of our pre-theoretic assumptions will

turn out to be false—among them, that ordinary objects are not states

of our minds! So it seems question begging to argue against the

identity phenomenalist reading by assuming that mental states cannot

be spatial.

I think the best reason to reject the identity phenomenalist

interpretation is that it is incompatible with the many of the very texts

that are used to motivate it.48 In many of the texts in which Kant

identifies appearances with (a species of) representation, he also

claims that representations are representations of appearances, i.e.

48 There is also the lingering problem of whether it is compatible with the Refutation of Idealism; see section 2.4. and part IV of Guyer (1987).

37

that representations are representations of objects, and those objects

are appearances. For instance,

[. . .] external objects (bodies) are merely appearances, hence also nothing other than a species of my representations, whose objects are something only through these representations, but are nothing separated from them. (A370-1; my emphasis)

everything intuited in space or in time, hence all objects of an experience possible for us, are nothing but appearances, i.e., mere representations, which, as they are represented, as extended beings or series of alterations, have outside our thoughts no existence grounded in itself (A490-1/B518-9; my emphasis)49

In both passages, Kant describes appearances as representations but

also as objects of representation. If this is correct, then Kant thinks

that the sense in which an appearance is a representation is

compatible with it being the objet of a representation. For instance,

the sense in which this table ‘is’ a representation is compatible with it

being the object of my perception of it. Assuming that the

representations that empirical objects ‘are’ are not always self-

representational (e.g. the table is not identical to a table-ish visual

perception that also represents itself), it follows that the objects

cannot be our representations of them. For instance, my visual

perception of this table cannot be this table, because my visual

perception does not represent itself. To make the identity

phenomenalist view consistent with the very texts that motivate it, we

need to ‘double’ our representations: a visual perception of the table

49 Allison (2004, 36) attempts to explain away the apparently phenomenalist implications of this passage.

38

and then the representation that the table is. But what could that

representation be? It must be present when and only when the table

exists, and my perception of the table must be intentionally directed

at it. While this is not conclusive, it is good evidence that the identity

phenomenalist interpretation should be abandoned.

Is there any way to free Kant from the apparent consequences of

his tendency to identify appearances with representations of them?

One standard strategy is to say that Kant is simply being sloppy: he

means that phenomena are the objects of our representations, not

that they literally are those representations.50 This is implausible,

however; the passages in question occur throughout the Critique, in

both editions, and they remain after Feder-Garve pointed out their

apparently phenomenalist implications. On the other hand, their

persistence in the B Edition suggests that they do not, and never were

intended to, commit Kant to a form of identity phenomenalism. How

could Kant claim Feder-Garve had misunderstood him if he had

identified appearances with representations? This suggests that

another reading is possible, but does not tell us what it is.

One strategy would be to claim that Kant does not mean the ‘is’ of

identity, but the ‘is’ of grounding. Sometimes, apparent claims of

50 Van Cleve (1999) refers to the “act-object (‘ing’-‘ed’) ambiguity of words like ‘representation’ (which is also possessed by words like Vorstellung in German)” (7). However, I’m not sure my linguistic intuitions (about either German or English) agree with Van Cleve’s here.

39

identity are really claims about grounding relations. For instance, if I

say ‘pain is C-fiber firing’ I might mean the type-identity thesis that

the state of being in pain is the state of C-fiber firing. But I might also

mean that all there is to pain is C-fiber firing, that if one is in pain it is

in virtue of C-fiber firing, or that C-fiber firing non-causally grounds

the state of being in pain. On this view, in claiming that appearances

are representations, Kant is claiming that the contents of

representations ground the existence and empirical properties of

appearances.

But this is not the plain meaning of the relevant passages. At A371

Kant claims that appearances are a species (Art) of representations;

while ‘is’ can be interpreted in a number of ways (e.g. the ‘is’ of

constitution), it is hard to interpret ‘As are a species of Bs’ in any

other way than: every A is a B, which means every A is identical to a B

(namely, itself). While I am sympathetic to the grounding

interpretation of these passages, I feel that there is something to

these texts that has not been explained (or explained away).

A third alternative, proposed by Wilfred Sellars, and which may

ultimately face the same problem, relies on the Cartesian distinction

between the formal and objective reality of representations (in

Cartesian terminology, ideas).51 The objective reality of an idea is the

51 Sellars (1968), ch.2 and (1976). See also Cummins (1968) and Aquila (1979) and (1983), ch. 4. Descartes discusses the formal and objective reality of ideas in Meditation three.

40

representational character of the idea, its character as a

representation with a certain content. Consequently, we can talk

about the object of an idea without assuming that there is an object

‘external’ to the idea; to talk of the ‘internal’ object of the idea is just

to talk about that idea’s objective reality. For instance, we can

coherently talk about God without presupposing that God exists

‘outside’ our idea of him; this God-talk is to be understood as talk

about our idea of God in its objective realty, i.e. to talk about the

content of our God-idea. Translating this back into Kant, we might

take his claims that appearances are representations as claims to the

effect that appearances are representations considered in their

objective reality, or, in other words, that talking about appearances,

objects of representations, is just talking about representations and

their contents.

There are at least two problems with this strategy, however. For

one, it is arguably no less a distortion of the plain letter of the text

than the other interpretations. If Kant meant that appearances are

representations considered with respect to their objective reality why

didn’t he simply say that, rather than stating that they are a species of

representations? Secondly, it is surely not correct that, on Kant’s

view, talk about appearances is equivalent to talk about the objective

reality of representations. Kant is not attempting a semantic analysis

of appearances in terms of representations. It is more plausible to

41

read Kant as claiming that appearances are grounded in

representations (non-semantically) and their objective reality

(content). So this proposal may collapse into the previous one.

Finally, it requires determining the exact nature of that grounding

relation. In doing so, we need to bear in mind the following question:

is there enough metaphysical distance between appearances and the

contents of representations that we are violating Kant’s dictum that

appearances are representations (considered in respect of their

objective reality)? If so, that is a sign that we are on the wrong track

in our interpretation.

3.2 Qualified phenomenalism

Kant repeatedly claims that our representations alone do not ground

the existence of their objects. At A92/B125 he writes that

“representation in itself does not produce its objects in so far as

existence is concerned” and in a 1792 letter to J.S. Beck he dismissed

the Feder-Garve interpretation with one line: “I speak of ideality in

respect of the form of representation, while they construe it as

ideality in respect of the matter, i.e. ideality of the object and its

existence” (Ak. 11:395). I take the first passage to mean that the

existence of empirical objects is not wholly grounded in the contents

of our experience; something else must be added. I take the second

42

passage to mean that Feder and Garve misattributed to him the

opposite view: that all there is to the existence of an object in space is

our having mental states with a certain content. But all this shows is

that strong phenomenalism is not Kant’s view. It leaves open the

possibility that he accepts qualified phenomenalism: the existence and

core physical properties of objects in space is grounded in the

contents of our representations of them.

The first question to be answered is, what, in addition to the

contents of our representations, grounds the existence of empirical

objects? The natural answer, for the qualified phenomenalist, is that

there must be things in themselves that appear as these objects! Kant

repeatedly insists that it is a conceptual truth that appearances are

appearances of something that is not itself an appearance, a thing in

itself. For instance,

In fact, if we view the objects of sense as mere appearances, as is fitting, then we thereby admit at the very same time that a thing in itself underlies them, although we are not acquainted with this thing as it may be constituted in itself, but only with its appearance, i.e. with the way in which our senses are affected by this unknown something. Therefore the understanding, just by the fact that it accepts appearances, also admits to the existence of things in themselves, and to that extent we can say that the representation of such beings as underlie all appearances, hence of mere intelligible beings, is not merely permitted but also unavoidable. (Prol, Ak. 4:314-5; Au’s emphasis)52

On the qualified phenomenalist reading, this means that the existence

of an appearance requires (a) a representation of an object, and (b) a

52 See also A251-2, Bxxvi-xxvii, B306, and B307.

43

thing in itself that appears as that object. A fully developed qualified

phenomenalist reading would require saying precisely what it means

for a thing in itself to appear as an empirical object (an object of

experience), but for reasons of space I can only sketch an answer

here. At the least, the qualified phenomenalist might require that the

thing in itself causally affect the experiencing subject, and that the

sensory content thus produced be involved in the experience of the

appearance. Some scholars have suggested that the properties of

appearances are structurally isomorphic to the properties of things in

themselves, but I will not pursue that idea here.53

The qualified phenomenalist also owes an answer to the question,

which are the representations whose content (partly) grounds the

existence of empirical objects and (wholly) grounds their core physical

properties? The natural answer is ‘experience,’ so the qualified

phenomenalist owes us an interpretation of what Kant means by

‘experience,’ what its content is, and how it grounds (partly) the

existence and empirical properties of appearances.54 We have already

seen that, for familiar reasons, Kant cannot ground the existence of

empirical objects in our mere perceptions of them: sometimes we

53 E.g. Findlay (1981), 92-93. See also Van Cleve (1999), 155-162.54 Passages that support the grounding of the existence of objects in experience include A245/B276, B279, A490-1/B518-9, A520/B492-A521/B493, A494/B522. Aquila (1983) develops a detailed textual case that appearances exist in being experienced. For a non-phenomenalist reading of these passages see Allison (2004), 40 and 41.

44

misperceive objects, objects exist while unperceived, and there are

objects we cannot ever directly perceive.

Kant distinguishes experience from perception in the A Deduction,

writing:

There is only one experience, in which all perceptions are represented as in thoroughgoing and lawlike connection, just as there is only one space and time, in which all forms of appearance and all relation of being or non-being take place. If one speaks of different experiences, they are only so many perceptions insofar as they belong to one and the same universal experience. The thoroughgoing and synthetic unity of perceptions is precisely what constitutes the form of experience, and it is nothing other than the synthetic unity of the appearances in accordance with concepts (A110).

In this sense of experience there is only one experience. I take this to

mean that, inter-subjectively, there is only one experience as well: my

perceptions and your perceptions are only ‘experiences’ to the extent

that they cohere with the one universal experience. I will dub this

sense of experience ‘universal experience.’ Kant, in this passage,

does not tell us much about what universal experience is, or what its

contents are. He does tell us what it is composed from perceptions,

that it has an a priori form (space, time, and categories), and that the

perceptions that constitute it are in “thoroughgoing and lawlike

connection.”

Elsewhere, he sheds further light on the coherence relation that

defines universal experience:

In space and time, however, the empirical truth of appearances is satisfactorily secured, and sufficiently distinguished from its kinship with dreams, if both are correctly and thoroughly connected up according to empirical laws in one experience. Accordingly, the

45

objects of experience are never given in themselves, but only in experience, and they do not exist at all outside it. (A493/B521)

I take this to mean that perception Pn coheres with perceptions P1

through Pn-1 to the extent that the causal laws observed in P1 through

Pn-1 are observed in Pn. This gives us reason to exclude hallucinatory

perceptions from universal experience: hallucinatory perceptions

involve apparent violations of the causal laws that are observed to

hold in our ‘waking’ perceptions, so they do not cohere with those

other perceptions.

We know a priori something very general about the form of

universal experience, of course: it will be spatiotemporal and the

principles of experience (applications of the categories) will hold in it.

But that does not determine the determinate a posteriori content of

universal experience, and the idea of a qualified phenomenalist

analysis of empirical objects is to hold that their existence and

empirical properties are grounded in that fully determinate a

posteriori content. So we might suggest the following analysis of

experience:

(Experience) Universal experience consists in the largest internally coherent subset of perceptions that obeys the principles of experience. A subset of perceptions is internally coherent to the degree to which causal regularities hold among its contents.

On a qualified phenomenalist reading of Kant, this might be taken as

the representation whose content (partly) grounds the existence and

(wholly) the ‘core physical properties’ of spatiotemporal objects.

46

However, I think there are at least two problems with this analysis of

universal experience:

(i) Unperceived objects. Kant holds that there are

spatiotemporal objects we cannot perceive. This by itself

would pose a problem for the proposed definition of

experience, since, on the qualified phenomenalist view, that

definition entails that there cannot be unperceived

spatiotemporal objects. But Kant further claims that we can

experience unperceivable objects through perceiving their

effects and inferring their existence from causal laws. So the

definition of experience needs to be refined.55

(ii) Secondary qualities. As we discussed above in the section on

Kant’s empirical realism, Kant distinguishes between the

properties spatiotemporal objects actually have ‘in

themselves’ and those they merely appear to have in sense

perception. He has a basically Lockean distinction between

primary and secondary qualities at the empirical level. Since

we perceive objects as having secondary qualities, the

definition of Experience given above, combined with the

phenomenalist analysis, entails that empirical objects have

secondary qualities. We need to further refine our definition

55 A226/B273. Langton (1998), 186-190.

47

of experience to eliminate secondary qualities from the

empirical properties of objects.

We need to refine the conception of experience so as to include

unperceived objects and exclude secondary qualities. This might push

us towards a more ‘scientistic’ conception of Kantian experience, on

which experience is something like the ideal scientific theory of

objects in space and time.56 The form of that theory is a priori

determinable from the forms of experience: it will represent persisting

substances in a 3-D Euclidean space obeying universal casual laws

and in simultaneous mutual interaction. However, the determinate a

posteriori content of that theory will be grounded in the perceptions

subjects actually have. Here is a first gloss on such a definition of

experience:

(Experience) Universal experience is the maximally unified and lawful representation of objects in space and time that is compatible with the a priori forms of experience and justified by the totality of subjects’ perceptual states, or the conjunction of such representations if there is no unique such representation.57

56 This interpretation of experience bears some resemblance to that given in Cohen (1871), without Cohen’s Neo-Kantian reading of the thing in itself as the unapproachable limit of scientific knowledge.57 Conjunction is usually defined for sentences, but it can be easily generalized to all representations that have correctness conditions: if A and B are representations in some mode m, the representation A&B is the representation in mode m (if there is one) that is correct just in case A is correct and B is correct.

48

To fully develop such a view, a lot more would have to be said about

exactly how the content of experience is grounded in and justified by

the contents of subjects’ perceptual states, but this gloss is enough to

give us a sense of what a developed phenomenalist reading of Kant

would look like.

Before continuing, though, I’d like to make a few clarifications

about this proposed qualified phenomenalist interpretation:

That universal experience is a representation merely means that

it has a content; it represents objects as being a certain way. In

what follows, I will sometimes refer to it as a ‘theory.’ I do not

mean this term to bring with it any anachronistic connotations; I

mean merely that it will be representation with a content more

complex than any individual perceptual experience or judgment.

By ‘lawful,’ I mean that universal experience will represent

empirical objects as obeying deterministic, universal causal

laws. This much is clear from the passages quoted already. By

‘maximally systematic’ I mean that universal experience will, as

best as it can, represent the world as governed by a system of

laws that have the form of a logical system: relatively specific

lower-level laws subordinated to higher-level laws of greater

generality, etc.

The sensory states that justify universal experience are the

totality of all sensory states of all human subjects across all

49

times. Consequently, the content of universal experience

correctly understood, does not change. What changes is our

collective understanding of the content of universal experience.

3.3 Criticisms of Phenomenalist Readings

Since the Feder-Garve objection to Kant has been around almost as

long as the Critique itself, there have accumulated many objections to

broadly phenomenalist readings of Kant’s idealism. Perhaps the most

comprehensive list of such objections is given by Allais (2004). Her

objections specifically to a phenomenalist reading of Kant include:

“Kant’s claim that his notion of appearance implies that there is

something that appears.” I have already explained how the

qualified phenomenalist can accommodate this point.

“Empirically real objects and the space they inhabit are public.”

Allais seems to assume that, on a phenomenalist analysis of

objects in space, objects are ‘private,’ meaning the objects each

subject perceives are constituted by sense data of that subject,

which, by definition, cannot be perceived by other subjects. But

the qualified phenomenalist conception of universal experience

sketched above is explicitly ‘non-private’ in this sense; it is

based on the perceptions of all subjects.

50

“Kant’s distinction between primary and secondary qualities.”

However, the qualified phenomenalist can claim that while our

perceptions represent objects as having secondary qualities, the

best scientific theory justified by the totality of those

perceptions (universal experience) does not represent them as

having those properties, because there is a better theory

available: objects do not possess such properties but do possess

powers to cause us to perceive them as having such properties.

There is no in principle barrier to a qualified phenomenalist

allowing the distinction between primary and secondary

qualities.

“Kant’s realism about the unobservable entities of theoretical

science.” Kant is a scientific realist, in that he accepts the

existence of unobservable entities posited by our best physical

theories (magnetic matter, Newtonian ‘lamellae’). Allais thinks

this is incompatible with a phenomenalist reading, but it is

compatible with the conception of universal experience

developed in the previous section. If universal experience has

the content of the scientific theory best justified by our

perceptions, then universal experience can represent

unobservable (=unperceivable) objects.58

58 Cf. Allison (2004, 46) who also objects that phenomenalism is incompatible with Kant’s empirical realism.

51

“Empirically real objects exist through time and unperceived,

and are in causal relations.” The fact that empirically real

objects exist through time while unperceived might be thought

to pose a problem for phenomenalism, although it should be

remembered that Berkeley (at least on some readings) is a

phenomenalist and yet accepts that objects exist while

unperceived (although he denies that they stand in causal

relations). So it isn’t clear why Allais think this is incompatible

with phenomenalism. And it clearly is compatible with the

conception of universal experience developed in the previous

section: experience represents objects as existing through time

and unperceived, because a theory that represents them as

existing only when perceived would be far less unified and

lawful.

“We do not know what ideas are in themselves.” Allais’s idea

seems to be that the phenomenalist is committed to grounding

empirical objects not in ‘empirical ideas’ (temporally ordered

mental states available in conscious introspection) but in

‘noumenal ideas’ (the non-temporal states of the subject in

itself). But it is unclear why; the phenomenalist conception of

experience developed in the previous section explicitly grounds

52

the a posteriori content of universal experience in ‘empirical’

ideas, the totality of subjects’ perceptions.59

Allais appears to have conflated phenomenalist readings of Kant in

general with the ‘strong’ phenomenalism (or even identity

phenomenalism) discussed in section two, and one, moreover, that

identifies experience with mere perception.

3.4. The Problem of Things in Themselves

No discussion of Kant’s transcendental idealism would be complete

without a discussion of F.H. Jacobi’s famous objection to the critique:

“without the presupposition of the [thing in itself] I cannot enter the

[critical] system, and with that presupposition I cannot remain in it.”60

Jacobi is referring to a number of quite serious problems for Kant’s

transcendental idealist theory. They do not disappear on other

interpretations, but they are especially serious for the traditional

phenomenalist reading. Unlike the problems we discussed earlier,

however, which were specifically problems for the phenomenalist

analysis of appearances, these problems, as Jacobi indicates, concern

59 Allais may have in mind the fact that perceptions are themselves appearances, and thus exist in virtue of the very universal experience they ground. If so, her objection is a form of the ‘problem of affection’ discussed in the next sub-section, and treated more fully in Stang (ms).

60 Jacobi, Werke, vol. II, p. 304.

53

the thing in itself, and the relation between the things in themselves

and appearances.

3.4.1 The Unknowability of Things in Themselves

Kant is committed to both of the following theses:

(Existence) There are things in themselves.

(Humility) We know nothing about things in themselves.

While these are not, strictly speaking, incompatible, they are in

tension, for Humility appears to remove any warrant Kant might have

for asserting Existence.

On the qualified phenomenalist view developed in the previous

section, it is part of the concept of appearance that the object exists in

virtue of an experience which results from causal affection by a thing

in itself. On this reading, Kant would be justified in claiming that if

objects are appearances, then there are things in themselves of which

they are appearances. But what justifies him in claiming that objects

are appearances, if the concept of appearance has this requirement

built into it? It appears that Kant has asserted the existence of mind-

independent objects, thing in themselves, an assertion which violates

the principles of his own epistemology.

54

But it gets worse for the traditional view. Kant does not merely

claim that things in themselves exist, he also asserts that,

(Non-spatiality) Things in themselves are not in space and

time.

(Affection) Things in themselves causally affect us.61

Many of Kant’s early readers concluded that Kant’s philosophy is

inconsistent: he claims that we cannot know the very assertions he

makes about things in themselves. Kant’s own theory renders itself

unsayable. 62

I do not want to suggest that each of these three problems -- how

to square Humility, with Non-Spatiality, Affection, and Existence -- are

on a par. Since Non-spatiality makes only a negative claim, it may be

61 It is not universally granted that Kant accepts the Affection claim; Vaihinger (1887) lists Fichte, Beck, Maimon, and Heramnn Cohen, among others, as “Kantianer” who denied noumenal affection. In the third chapter of his (1924), Erich Adickes assembles an impressive array of textual evidence that Kant did accept Affection. See especially A190/B235, A387, A494/B522, Ak. 4:289, 4:314, 4:318, 4:451 and 8:215.

62 Van Cleve points out that p and It is unknowable whether p are logically consistent but asserting both might constitute a pragmatic self-contradiction (1999, 135). But, as Hogan (2009a), 61n3) points out, Kant also claims that it is “indubitably certain” (A48) that things in themselves are not in space and time. Ameriks (2003, 29-30) argues that Existence and Affection constitute a pre-theoretic belief that Kant never has reason to deny (although he does successively give up his pre-theoretic belief that these beings are spatial, etc.), to which Hogan correctly counters that Non-spatiality is surely not such a pre-theoretic assumption (2009a, 50-51) and that, even if Affection is a pre-theoretic belief that is never defeated by the critical epistemology, it is undermined by the discovery that none of the objects we experience are things in themselves (2009b, 503).

55

easier to make it consistent with Humility. For instance, at B149 Kant

writes: “it is not yet a genuine cognition if I merely indicate what the

intuition of the object is not, without being able to say what is then

contained in it.” This suggests that, while Kant’s usually unqualified

statements of our ignorance of things in themselves (they are “not

cognized at all” A30/B45), his considered view might be more

qualified: we know nothing of the positive properties of things in

themselves.63 And perhaps some strategy like the one sketched above

can be made to work: it is a conceptual truth that if there are

appearances, there must be something that is not itself an

appearance, and this thing is a thing in itself. But Affection looks

especially difficult to square with Humility.64

3.4.2 Things in themselves as causes

The issue of things in themselves affecting us raises another problem

for Kant’s theory, for Kant also argues that categories like cause-

effect cannot be meaningfully applied to things in themselves.

Without an intuition “[the category] has no sense, and is entirely

63 Langton (1998) defends such an interpretation, on which we lack knowledge specifically of the intrinsic properties of things in themselves.

64 See Hogan (2009a) and Stang (2012b).

56

empty of content” (A239/B298). Since things in themselves cannot be

intuited, categories (including cause-effect) have no sense or content

when applied to things in themselves. Jacobi and others thought this

was yet another inconsistency in Kant’s philosophy: he denies that

categories can be applied to things in themselves, but then he applies

the category cause-effect to them!

However, one has to be careful in interpreting Kant’s denial of

‘sense’ or ‘meaning’ to categories as applied to thing in themselves. It

is tempting to read this as meaning that the thought of things in

themselves falling under categories is literally nonsense, but the

textual evidence shows that Kant is making a weaker point: thinking

of things in themselves under the categories has no cognitive sense,

i.e. in making such judgments we do not cognize anything. For

instance,

[. . .] the categories are not restricted in thinking by the conditions of our sensible intuition, but have an unbounded field, and only the cognition of objects that we think, the determination of the object, requires intuition; in the absence of the latter, the thought of the object can still have its true and useful consequences for the use of reason [. . .] (B166n)65

We can think of any objects whatsoever using the categories. In fact,

this is unavoidable; the categories are the most basic concepts for

65 Cf. B166-7, A88/B120, A254/B309. In the remainder of this passage, Kant refers to the practical use of the thought of categories without intuitions; this is a reference to the crucial role that things in themselves play in his theory of freedom. See Critique of Practical Reason, Ak. 5:43, 54-55.

57

thinking of objects in general, so we cannot think about anything

whatsoever without using some categories to do so. But in thinking

about the things in themselves using categories we do not thereby (a)

know that there are things in themselves falling under the categories

or (b) even that it is possible for there to be things in themselves

falling under the categories.66 On balance, therefore, I think the

strongest form of Jacobi’s objection—that Kant’s view entails that the

categories cannot be applied, even in thought, to things in themselves

—rests on a misunderstanding.67 This still leaves, though, the

pressing problem of, given Kant’s Humility doctrine, he could have

any epistemic warrant for making the various substantive claims he

does about things in themselves (Existence, Non-spatiality, Affection).

3.4.3 The Problem of Affection

Jacobi raises yet another further problem about Kant’s theory of

experience. He notes Kant’s definition of sensibility as the capacity

“to receive representations through the manner in which we are

affected by objects” (A19/B33) and poses a dilemma: are the objects

66 This is not to deny (or to affirm) that Kant holds that we can know that things in themselves exist, or that they affect us; it is only to say that merely in thinking there are existing things in themselves or things in themselves are the causes of my sensory states I come to know anything.

67 Cf. Van Cleve (1999), 137; Adams (1997), 820-1.

58

that affect our sensibility appearances or things in themselves? They

cannot be appearances, Jacobi argues, because that would involve

applying the categories to things in themselves. And they cannot be

appearances, because appearances exist in virtue of the very

experiences they are (allegedly) causing. He concludes that Kant’s

system is inconsistent.68

We have already discussed, and rejected, the argument of the

second horn of Jacobi’s dilemma: we can think but not know that

things in themselves causally affect us. But what about the first horn?

Hans Vaihinger explains the argument well:

Or one understands by affecting objects the objects in space; but since these are only appearances according to Kant, and thus our representations, one falls into the contradiction that the same appearances, which we first have on the basis of affection, should be the source of that very affection.69

‘First’ here does not refer to temporal priority, but to metaphysical

priority: if p is true in virtue of q, then q is ‘prior’ to p. Jacobi and

Vaihinger assume that appearances exist in virtue of the contents of

our experience of them:

(Trans. Idealism) If x is an appearance, then x exists in virtue of the fact that subjects experience x.

If we are empirically affected, though, it follows that:

68 Jacobi, Werke, vol. II, 291-310. Fichte raises the same objection in the Second Introduction to the Wissenschaftslehre; cf. Fichte, Werke I, 488.

69 Vaihinger (1887), vol. 2, 53.

59

(Empirical affection) For some x, x is one of the causes of subjects’ experience of x.

For instance, this computer is one of the causes of my current

experience of it. But these assumptions are inconsistent if we assume

the following plausible principle:

(Exclusion) If x exists in virtue of the fact that p, then x cannot be even a partial cause of the fact that p.

Intuitively, this principle says that no object can be even a partial

cause of the very fact in virtue of which it exists; if it were, it would be

a partial cause of its own existence. In the context of Kant’s theory of

experience, it means that appearances cannot ‘reach back’ and cause

the very experiences in virtue of which they exist. From the 1780s

until today, many have taken this problem to be fatal to Kant’s theory

of experience.70

Before concluding this section, I want to briefly sketch how the

qualified phenomenalist interpretation from earlier could solve

Jacobi’s problem. The qualified phenomenalist can reinterpret the

principles from above as:

70 It is sometimes misleadingly referred to as the problem of double affection. Double affection is the doctrine that accepts both horns of Jacobi’s problem, famously defended by Adickes (1924) and (1929). Jacobi’s problem only arises on the first horn: empirical affection. Van Cleve (2008, 165-6) rejects the doctrine of empirical affection for this very reason. Stang (forthcoming) attempts to solve the problem. For more on the doctrine of double affection see Drexler (1904), Adickes (1920), Kemp Smith (1962, especially Appendix C), Weldon (1958), Vaihinger (1892) vol. 2, 35-55, esp. 53-55, and Gram (1975).

60

(Trans. Idealism*) If x is an appearance, then x exists partly in virtue of the contents of universal experience.

(Empirical Affection*) If x is an appearance, then x is a partial cause of subjects’ perceptions of x.

The conjunction of these two claims is consistent, not only with the

Exclusion principle from earlier, but with this strengthened version:

(Exclusion*) If x exists partly in virtue of the fact that p, then x is not even a partial cause of the fact that p.

This suggests that the path to a consistent qualified phenomenalism

that is compatible with empirical affection and noumenal affection lies

in distinguishing perception from ‘universal experience,’ making the

former the relata of empirical affection and making the latter the

ground of empirical objects’ existence and empirical properties, as I

did in the previous section. But I do not have the space here to

develop such a view more fully.71

5. The ‘Dual Aspect’ View

Because the phenomenalist interpretation of transcendental idealism

held such sway, not only among Kant’s contemporaries, but for

generations of German philosophers as well, these problems for the

phenomenalist construal of transcendental idealism were taken to be

71 See Stang (ms).

61

evidence that Kant’s view itself is inconsistent.72 In the twentieth

century, the phenomenalist (or ‘Berkeleyan’) interpretation of

transcendental idealism is associated with P.F. Strawson, whose

massively influential (1966) argued that, for many of the reasons we

have seen, transcendental idealism was a blunder on Kant’s part.73

However, Strawson claimed, the core arguments of the Critique do

not in fact rely on it and could be reconstructed independently of it.

In the 1960s and 1970s a group of scholars, in some cases in direct

opposition to Strawson, developed a non-phenomenalist, anti-

metaphysical reading of transcendental idealism, the ‘dual aspect’

view.74 These scholars took the textbook problems for phenomenalism

(especially, the problem of affection) as evidence that this was the

wrong interpretation of Kant’s position to begin with. They sought to

rescue transcendental idealism from what they took to be the

phenomenalist misconstrual, defend its philosophical cogency from its

detractors, and show, contra Strawson, that the central arguments of

72 I do not want to suggest that everyone read the Critique this way until the twentieth century. Already in the nineteenth century, the Neo-Kantian movement developed a non-phenomenalist interpretation of Kant, but that lies outside the scope of this survey. See especially Cohen (1871/1885), (1907); as well as Köhnke (1986), Patton (2005), and Richardson (2003).

73 Strawson (1966), 16, 38-42, 253-73.

74 I have in mind mainly Bird (1962), Prauss (1974), Allison (1974), (1976), (1983) and (2004). Cf. Collins (1999), Dryer (1966), ch. 11; Matthews (1969). Earlier ‘dual aspect’ readings include Paton (1936), vol. 1, 61, and Beck (1960).

62

the Critique do rely on transcendental idealism. This was as much a

philosophical defense of Kantian transcendental idealism as it was an

interpretive-exegetical project.

They developed what has become known as the ‘dual aspect’ view.

They argue that many of the classic problems for the phenomenalist

reading (e.g. affection) arise because it was mistakenly assumed that

appearances and things in themselves are distinct kinds of objects.

They argued instead that the appearance/thing in itself distinction is

not an ontological distinction between two kinds of objects, but an

adverbial distinction between two different perspectives or stances we

can take on one and the same set of objects: we can consider them as

they appear, or as they are ‘in themselves.’

4.1 One Object, Not Two

In numerous passages, Kant describes the appearance/thing in itself

distinction, not as a distinction between two different objects, but as a

distinction between two ways of considering one and the same object.

For instance,

[. . .] the same objects can be considered from two different sides, on the one side as objects of the senses and the understanding for experience, and on the other side as objects that are merely thought at most for isolated reason striving beyond the bounds of experience. (Bxviii-Bxix, note)

63

[. . .] the reservation must well be noted that even if we cannot cognize these same objects as things in themselves, we are lat least able to think of them as things in themselves. (Bxxvi)75

The general characteristic of such passages is that they use the same

chain of pronouns to refer both to appearances and things in

themselves. This strongly suggests that one and the same object can

be an appearance and a thing in itself, or, to put it another way, the

distinction between appearance and thing in itself is not a distinction

between two or more objects, but a distinction between two different

aspects of, or ways of considering, one and the same object. One and

the same object can be considered as it appear to us in experience, or

as it is in itself. Considered in the former way, the object must

conform to our a priori intuitional forms, so it is in space and time.

Considered in the other way, the object may not be in space and time.

Some ‘dual aspect’ readers cite the increased frequency of such

passages in the Prolegomena and the B Edition as evidence that Kant,

realizing that his distinction between two aspects of objects was being

conflated with a distinction between two kinds of objects, sought to

remedy this interpretation by emphasizing precisely this point.

Prauss (1974) notes that, in most cases, Kant uses the expression

75 See also A35/B51, B69, B306, A360. The Dual Aspect view is more or less directly stated by Kant in the Opus Postumum (Ak. 22:43-44). However, since some scholars think that Kant revised fundamental aspects of his Critical theory in that later work (the idea of a ‘post-Critical’ Kant), this is not as decisive evidence as it would otherwise be.

64

‘Dinge [Sachen, Object, Gegenstand] an sich selbst’ rather than the

shorter form ‘Dinge an sich.’ He argues that ‘an sich selbst’

functions as an adverb to modify an implicit attitude verb like ‘to

consider’ [betrachten]. He concludes that the dominant use of these

expressions is as a short-hand for “things considered as they are in

themselves.”76

Different scholars understand this distinction in different ways.

The main difference is between epistemological and metaphysical

‘dual aspect’ interpretations.77 On the epistemological reading, the

distinction between appearances and things in themselves is simply a

distinction in the standpoint from which we consider them. We can

consider objects as objects of knowledge for discursive spatiotemporal

cognizers like us, in which case we are considering objects as

appearances. Or we can abstract from our particular cognitive

conditions and consider objects merely as objects for a mind in

general, in which case we are considering them as things in

themselves. It is crucial to the epistemological reading that there is

no sense in which the ‘transcendental’ perspective on objects as

things in themselves gets at how objects ‘are in themselves.’ The

point of Kant’s transcendental idealism, epistemological interpreterss

stress, is to get away from the incoherent idea of a ‘view from

76 Prauss (1974), 14-15.

77 Allison (2004), 52.

65

nowhere’ in which we could know objects as they ‘really are in

themselves.’78

By contrast, metaphysical identity interpreters take the distinction

to carry more metaphysical weight. They interpret the

appearance/thing in itself distinction as a metaphysical distinction

between two different classes of properties had by objects, for

instance, their relational properties and their intrinsic properties.

Appearances are objects qua bearers of ‘empirical properties’ (e.g.

relational properties) while things in themselves are the very same

objects qua bearers of ‘noumenal’ or ‘non-empirical’ properties (e.g.

intrinsic properties). The next two sub-sections explore the

epistemological interpretation of Henry Allison. The remainder of the

sections concerns metaphysical ‘dual aspect’ readings, focusing on

the widely discussed interpretation of Langton (1998).

4.2 Allison’s ‘Epistemic’ reading

In modern Kant scholarship, the epistemic reading was first put

forward by Gerold Prauss, Henry Allison, and Graham Bird. Since

Allison’s work was most influential among English language

78 There is thus a natural connection between the epistemological interpretation of transcendental idealism and Putnam-style ‘internal realism,’ a point noted by Allison (2004), 454 note 17. The relation between transcendental idealism and contemporary anti-realism is explored in Allais (2003) and Van Cleve (1999), ch. 12.

66

scholarship, and most likely to be known to readers, I will focus on the

interpretation of transcendental idealism in Allison (1983) and the

revised and enlarged second edition, (2004). Allison’s writings

contain several distinct (and not obviously equivalent) formulations of

transcendental idealism. In this section I will concentrate on

reconstructing what I take to be the ‘core’ of Allison’s reading: his

interpretation of appearances and things in themselves, and his

reconstruction of the argument for the non-spatiality of things in

themselves.

The core insight of Kant’s epistemology in general, and his

transcendental idealism in particular, according to Allison, is the

principle that we possess a discursive intellect. A discursive intellect

is one that passively receives representations of particular objects and

then spontaneously subsumes those represents objects under general

concepts; consequently, a discursive intellect must possess a sensory

faculty (through which it receives sensory data and intuits individual

objects) and a conceptual faculty (through which it form s general

concepts and applies them to objects).79 By contrast, an intuitive

intellect brings into existence its objects merely by representing them,

and thus has no need to receive representations of objects from

outside.80 But that is not all there is to the discursive nature of our

79 A50-1/B74-5.

80 For Kant’s conception of an ‘intuitive intellect’, see CPJ Ak. 5:406. He discusses the related (though not necessarily identical) notion of a

67

intellect, Allison argues.81 Kant’s key insight is that our sensible

faculty has its own epistemic conditions.

An ‘epistemic condition’ is Allison’s term for a representation we

must apply to objects in order to cognize them.82 Space and time are

epistemic conditions, as are the categories. If E is an epistemic

condition then necessarily if we know an object O, in knowing it we

represent it using E.83 Some of our epistemic conditions follow from

the general fact that we are discursive cognizers (the categories) and

some follow from the more specific fact that we are spatiotemporal

discursive cognizers (space and time). Representing objects using the

categories is an epistemic condition for any discursive intellect, i.e.

for any intellect that must conceptualize objects given passively in

non-sensible intuition at A249, A256/B311-2, B307, and A286/B342.

81 Strawson (1966, 20-21) interprets discursivity more minimally as the claim that we have singular representations of objects which we subsume under general concepts; Strawson thinks Kant went wrong in assuming that this ‘logical’ dualism must be explained by a ‘dualism’ of mental faculties: a receptive faculty of intuitions, and a spontaneous faculty of concepts. Allison (2004, 12-13) responds to Strawson on these points.

82 Allison (2004), 11, 14.

83 We have different epistemic conditions for different kinds of objects, e.g. space is an epistemic condition for outer objects, but not for inner intuitions. So the definition should really be: E is an epistemic conditions of objects on kind K iff necessarily in cognizing objects of kind K we represent them using E. However, in the body of the text I suppress this complication.

68

sensory intuition.84 So space and time are epistemic conditions of

spatiotemporal discursive cognition of objects, while the categories

are epistemic conditions of discursive cognition of objects in general.

Any discursive intellect must conceptualize sensibly intuited objects

using the categories, whether or not those objects are intuited in

space and time, or some other intuitional forms.85

This grounds a distinction between two ways of considering the

objects of our cognition. When we consider objects qua objects of our

cognition, we consider them as falling under the relevant epistemic

conditions. If E is an epistemic condition of cognition of objects, then

objects must fall under E (i.e. be accurately represented by E);

otherwise, in representing them with E, I would not be cognizing

objects but misrepresenting them. My representation of objects with

E would be an illusion, the very conclusion Kant wants to avoid with

respect to space and objects represented in space. This means that if

E is an epistemic condition of the specific kind of discursive cognition

of objects that we have, then E correctly represents those objects. So,

if space and time are the forms of our intuition, it follows that

empirical objects qua objects of the kind of discursive intellect we

84 “The pure concepts of the understanding are free from this limitation and extend to objects of intuition in general, whether the later be similar to our own or not, as long as it is sensible and not intellectual” (B148).

85 Allison (2004), 17.

69

have, are in space and time. But if we do not consider objects qua

objects of our specific kind of discursive intellect, but qua objects of

discursive intellect in general, we can no longer assume that our

specific intuitional epistemic condition still applies to them. The more

general epistemic conditions of all discursive cognition (in Kant’s

view, the categories) still apply to objects under this more abstract

perspective, however. So we can say that objects qua appearing

(objects of spatiotemporal discursive cognition) are in space, but qua

things in themselves (objects of discursive cognition in general) they

are not in space. This, in a nutshell, is Allison’s reconstruction of the

argument for the non-spatiality of things in themselves.86 While it is

legitimate to consider objects as things in themselves using the

categories, we do not thereby cognize them. This follows trivially

from the fact that space and time are epistemic conditions for us:

without representing objects in space and time, we can think of

objects using the categories, but those thoughts are not cognitions.87

4.3. Problems with the Epistemic reading

86 Allison also reconstructs Kant’s argument for the non-spatiality of things in themselves in the Transcendental Aesthetic (Allison 2004, 128-132), specifically, how Kant can respond to the famous ‘neglected alternative’ problem. However, that argument is independent of Allison’s larger interpretation of transcendental idealism, so I do not discuss it here; a ‘two object’ interpreter could, in principle, embrace that other argument. 87 Allison (2004), 18.

70

Allison’s interpretation has been challenged on a number of points by

other scholars. I discuss several of these objections in this section.

4.3.1 The Triviality Objection

Some scholars object that Allison’s reading of the non-spatiality

thesis, and the thesis that things in themselves are uncognizable by

us, renders it a tautology, a trivial logical consequence of definitions.88

I will represent the definition of ‘thing in itself’ talk (on Allison’s

interpretation) as follows:

(1) Things in themselves are F if and only if objects of discursive cognition as such are F.

And the non-spatiality thesis as:

(C) ~(Things in themselves are spatial)

But now the reader can see that to derive (C) from (1) we would a

further premise:

88 For the triviality objection see Guyer (1987), 336; Langton (1998), 8-12; Aquila (1980), 90; Van Cleve (1994), 4. For Allison’s own reply to the triviality charge, see his (2004), 18-19. Allison’s response is that the non-cognizability of things in themselves follows from a distinction, but the distinction is not trivial. This is confused, however. Distinctions do not have consequences; claims that distinctions do or not obtain have consequences. And, arguably, the statement that the appearance/thing in itself distinction obtains is a matter of definition (on Allison’s reading). In the body of the text I have attempted to reply on Allison’s behalf to the triviality objection.

71

(2) ~(Objects, considered, as objects of discursive cognition in

general, are spatial)

But this claim is not a definition, for it is equivalent to the claim that

the concept of a discursive cognition is more general than the concept

of a spatiotemporal discursive cognition, i.e. that a non-spatial

discursive intellect is conceivable. So although the non-spatiality of

things in themselves follows almost immediately from very general

truths, on Allison’s reconstruction, it is not correct to say that it is a

tautology, or that it is true by definition.

Nor is it true that the uncognizability of things in themselves is

trivial, on Allison’s reading. For that principle only follows from the

claim that there are sensible epistemic conditions, space and time.

And that, on Allison’s reconstruction, is the key insight that sets Kant

apart from both his rationalist and empiricist predecessors. Thus,

while Allison’s interpretation makes the argument for the non-

spatiality of things in themselves relatively easy, it does not render

the conclusion trivial.

4.3.2 Epistemic conditions entail realism

Robinson (1994) raises a quite general objection to Allison’s notion of

an epistemic condition, namely, an object must satisfy (fall under) a

representation if that representation is to constitute an epistemic

72

condition for that object.89 So in the claim that “objects qua

appearances” or “objects considered with our epistemic conditions”

the qualification “qua appearances” or “considered with our epistemic

conditions” is otiose. If space is an epistemic condition of outer

objects for us then this is because the objects we cognize are in space

simpliciter. The claim that objects are spatial because or in virtue of

space being an epistemic conditions entails either that these objects

exist in virtue of our representations of them (which results in

phenomenalism) or it entails that they are spatial in virtue of our

representation of them but would not be spatial otherwise, in which

case we are not cognizing them in representing them as spatial we

are misrepresenting them.90

Allison would reply to this objection, I take it, by pointing out that

it implicitly assumes that the claim empirical objects are in space is

coherent independently of specifying a perspective on those objects.

In the terminology of Allison (2004) it is committed to ‘transcendental

realism.’ I discuss the broad ‘meta-philosophical’ interpretation of

transcendental idealism below, in sub-section 5.4, but for now I want

to note that if this were Allison’s reply to the objection, then it would

show that the coherence of transcendental idealism, on Allison’s

reconstruction, rests on the premise that there is no coherent sense to

89 Robinson (1994) is a response, mainly, to Allison (1983) and (1987).

90 Robinson (1994), 420-22.

73

questions about how objects are independent of any perspective on

them. This is important, because it is not always clear that Allison’s

reconstruction does depend on this premise, and it is not clear where

Kant argues for such a conclusion.

4.3.3 Abstraction

One influential objection focuses on the role that ‘abstracting’ from

our spatiotemporal intuition plays in Allison’s reconstruction. Van

Cleve puts it somewhat facetiously:

How is it possible for the properties of a thing to be vary according to how it is considered? As I sit typing these words, I have shoes on my feet. But consider me apart from my shoes: so considered, am I barefoot? I am inclined to say no; consider me how you will, I am not now barefoot. (Van Cleve 1999, 8)91

To put the point less facetiously: if the object o, considered as an

object of spatiotemporal cognition, is spatial, then when we ascend to

a more general perspective, in which we consider o as the object of

discursive cognition in general, then we should not say that o is non-

spatial; we should merely not judge that it is spatial. On Van Cleve’s

reconstruction, Allison’s reasoning is as follows:

(1) o, qua object of spatiotemporal discursive intellect, is spatial.

91 See also Robinson (1987) and Guyer (1987), 279-329. Van Cleve (1999, 148-149) presents the argument in greater detail at p. and Allison responds in (2004, 42-45). As I understand Allison’s response to Van Cleve it is the same point as above: we should not assume that there is a way objects independently of an epistemic perspective on t hem.

74

(2) It is possible for there to be an object of discursive intellect that is not an object of spatiotemporal discursive intellect.

(3) o, qua object of discursive intellect in general, is not

spatial.

But (3) does not follow. All that follows is:

(3*) It is not the case that o, qua object of discursive intellect in

general, is spatial.

Determining whether Van Cleve’s objection is correct here requires

understanding in greater detail how ‘qua’ predications work in

Allison’s argument. I think Allison understands qua predication in

something like the following way:

(Qua) x qua F is G = it is a condition on considering x as an F that one represent it as G.

This analysis of qua predication would seem to vindicate the truth of

(1), on Allison’s interpretation. We can then reconstruct Allison’s

original reasoning as follows:

(1*) It is a condition on considering O as an object of spatiotemporal discursive cognition that one represent it as spatial.

(2*) It is not a condition on considering O as an object of discursive cognition in general that one represent it as spatial.

(3*) It is not the case that o qua object of discursive intellect in

general is spatial.

This way of analyzing qua predication would seem to vindicate Van

Cleve’s original objection to Allison’s argument: on Allison’s

75

reconstruction, Kant is not entitled to the claim that objects qua

objects of discursive intellect in general are non-spatial; he is at most

entitled to the claim that it is not the case that they are spatial.

But this is only a refutation of Allison’s interpretation if it is clear

from the texts that Kant claims (3) rather than (3*). But that is not at

all clear from the texts, for instance:

Space represents no property at all of any things in themselves nor any relation of them each other, i.e., no determination of them that attaches to objects themselves and that would remain even if one were to abstract from all subjective conditions of intuition. (A26/B42)

Prima face it is compatible with the letter of these texts that Kant is

claiming (3*) rather than (3). Note that (3*) is not the claim that we

cannot know, or justifiably assert that things in themselves are

spatial. It is the claim that it is false to say that they are spatial. The

relation of (3) to (3*) can be clarified through an empirical example of

abstraction: abstracting from differences among individual American

families and talking of ‘the typical American family.’ Assume that only

40% of American families own a dog. It would then be correct to say:

(4*) ~(The typical American family owns a dog).

But it would be false to say:

(4) The typical American family lacks a dog.

However, the sentence ‘the typical American family does not own a

dog’ is ambiguous between (4) and (4*). This sentence is analogous to

Kant’s claims that things in themselves are not spatial. Does this

76

mean the externally negated claim (3) or the internally negated claim

(3*)? Allison must read it as (3), which is not prima facie absurd.

The stronger objection to Allison’s view, as reconstructed here, is

that (3*) is too weak to be a plausible reconstruction of Kant’s non-

spatiality thesis. All that (3*) requires is that there is some

perspective on objects that is more general than the specifically

spatiotemporal form of cognition that we have. It does not even

require that it is possible that there be discursive intellects with a

non-spatiotemporal form of cognition. All it requires is that the

concept of discursive cognition as such is more general than the

concept of spatiotemporal discursive cognition, which, trivially, it is.

(3*) is compatible with it being impossible for there to be non-

spatiotemporal discursive cognition because all objects are

necessarily spatiotemporal and hence can only be cognized

spatiotemporally. In other words (3*) is compatible with

transcendental realism about space and time (as Kant defines that

term)!

As in the last section, I think Allison’s response to this objection

would be that it implicitly presupposes that there is a way objects are

independently of any perspective on them. In particular, the claim

that (3*) is compatible with all possible objects being spatial, and thus

cannot be a reconstruction of the non-spatiality thesis, begs the

question by assuming that that state-of-affairs does not need to

77

relativized to a perspective, e.g all possible objects as objects for a

certain kind of mind are spatial. Thus, I think that the coherence of

Allison’s reconstruction again depends upon the claim that there is no

‘standpoint-independent’ perspective on reality.

4.3.4 Things in themselves as more fundamental than

appearances

One major textual hurdle for Allison’s ‘epistemic’ reading of

transcendental idealism are the various passages in which Kant

describes things in themselves as more fundamental, more

ontologically basic, than appearances, or describes things in

themselves as the grounds of appearances. Allison appears to reverse

this relation of dependence because things in themselves (objects

from the relatively abstract transcendental perspective) are an

abstraction from appearances (objects from the more determinate

empirical perspective). Ameriks (1992, 334) raises this objection, and

Allison (2004, 45) replies to it. Allison does not offer an alternate

reading of the relevant texts, but instead points out that, in the case

where the relative fundamentality of the phenomenal and noumenal is

most important to Kant, namely the freedom of the will92, Ameriks

92 E.g. Kant’s claim in the CPR that our noumenal character causes our empirical character (A546/B574, A551/B579, A556/B584, A557/B585), and in Groundwork III that “the world of understanding contains the ground of the world of sense” (Ak. 4:453).

78

objection assumes, once again, that there is some fact of the matter as

to whether we are free or not, and this is to be settled by determining

whether we are free at the most fundamental level (the noumenal

level, on Ameriks’ reading). Once again, the coherence of Allison’s

reading rests on the premise that there is standpoint-independent

perspective on reality.

4.4 Transcendental Realism and Transcendental Idealism

Allison (2004) puts this ‘epistemic’ interpretation in the context of a

larger interpretation of transcendental idealism as a ‘meta-

philosophical’ position. Allison’s idea is that, since Kant seems to

regard transcendental idealism and ‘transcendental realism’ as not

only exclusive but exhaustive philosophical options, he must mean

something very general by transcendental realism. Consequently,

Allison argues, transcendental realists include not only the obvious

suspects (e.g. Wolff, Locke) but (more surprisingly) also Hume and

Berkeley. 93 Transcendental realism, as Allison characterizes it, is not

a discrete, statable thesis, but something more like what the later

93 In the second Critique Kant does identify Hume as a transcendental realist (Ak. 5:53); Beiser (2002), however, argues that this is because Kant knew Hume only from the first Enquiry and was thus unaware of the phenomenalist elements in the Treatise.

79

Wittgenstein would call a ‘picture’ to which pre-Kantian philosophy

was captive. Transcendental idealism, therefore, is not a discrete,

statable thesis either, but consists in freeing us from this ‘picture.’

Transcendental idealism is nothing less than a complete revolution in

our conception of what knowledge and philosophy are, an

“Umänderung der Denkart” as Kant describes it in the B Preface

(Bxvi).

Allison characterizes the transcendental realist ‘picture’ in at least

three ways, and it is not obvious that they are equivalent. First, he

characterizes transcendental realism as the very general thesis that

(what Allison calls) the ‘epistemic conditions’ of space and time exist

‘in themselves.’ I think Allison means instead that transcendental

realism is the implicit assumption that the question of whether space

and time exist ‘in themselves’ is a coherent one, because he regards

‘empirical idealism’ as itself a form of transcendental realism, and

Kant himself defines empirical idealism as the thesis that space and

time are mere illusions. The idea, I take it, is that taking seriously the

question do our sensible epistemic conditions (space and time)

accurately represent how reality is in itself? involves the mistaken

assumption that the notion of how reality ‘is in itself’ independently of

how we cognize it (i.e. independently of our epistemic conditions) is a

coherent one. This is Allison’s second main characterization of

transcendental realism: the assumption that there is a way reality is,

80

independently of a perspective on it.94 In the previous section I

argued that many of Allison’s reconstructions of particular Kantian

doctrines and arguments presuppose this conception of

transcendental idealism.

Thirdly, he characterizes transcendental realism as the implicit

commitment to the ‘theocentric’ paradigm of knowledge, the

assumption that God knows how object really are in themselves, and

that human knowledge is to evaluated by the extent to which it

matches up to that standard.95 Even atheists can be in the grip of the

‘theocentric’ model of knowledge, for they can still hold that human

knowledge is knowledge to the extent that it reveals how objects are

from the (in fact unoccupied) ‘God’s eye point of view.’

Transcendental idealism supposedly replaces the ‘theocentric’

paradigm with an ‘anthropocentric’ paradigm of knowledge through

recognizing (a) that we have epistemic conditions, (b) that they may

not be the only possible or actual epistemic conditions, and (c)

realizing that there is no sense to the question of whether ‘an sich’

reality matches those epistemic conditions. On the transcendental

idealist conception of knowledge, knowledge consists in objects

satisfying our epistemic conditions; our beliefs about the world do not

94 Allison (2004), 48.

95 Allison (2004), 28.

81

have to live up to some fictional standard ‘God’s eye point of view’ to

constitute knowledge.

However, all three characterizations of transcendental idealism face

significant problems, both philosophical and historical. First of all, it

is simply not true that, according to Kant, transcendental realism and

transcendental idealism are exhaustive options.96 He does famously

claim that “it is really this transcendental realist who afterwards plays

the empirical idealist” (A369) by which he means that, having

concluded (correctly Kant thinks) that we could never infer from our

inner states to the existence of transcendentally external objects in a

way that would secure knowledge of those objects, the transcendental

realist, who thinks that if there are objects in space and time then

they transcendentally external, concludes that we know nothing of

them (problematic idealism). But this means (at most) that

transcendental realism entails (or naturally leads to) problematic

empirical idealism; it does not entail that empirical idealism as such is

a form of transcendental realism. Dogmatic idealists (like Berkeley,

on Kant’s misreading of him) are transcendental realists in the

attenuated sense that they think that if there were objects in space,

they would be transcendentally external objects. But Kant by his own

96 Allison (2003), 23. The only evidence Allison gives for this claim are some texts (quoted earlier) in which Kant calls transcendental realism the ‘common prejudice’ and a passage from his notes for the What real progress? essay in which Kant claims that all philosophy before the Critique was essentially the same (Ak. 20:287, 377).

82

lights is a transcendental realist in this sense, for he thinks that things

in themselves are transcendentally external (ausser) objects!

Secondly, if transcendental idealism is equivalent to the thesis that

there is no standpoint independent perspective on reality, Allison

owes us a reconstruction of Kant’s argument for that (incredibly

strong) thesis. But Allison never gives this argument. The argument

from the discursive nature of our intellects to the claim that objects,

considered as they are in themselves (abstracting from the specifically

spatiotemporal nature of our intuition), is not such an argument.

Kant, after all, takes the Critique to establish the truth of

transcendental idealism, not merely to assume it.

Finally, regarding the idea of a ‘theocentric’ paradigm of

knowledge, Kant himself repeatedly contrasts our discursive form of

cognition with the intuitive intellect that God might possess; Kant

holds that God would, but we do not, cognize things as they are in

themselves.97 This does not mean that knowledge for human beings

consists in approximating this divine model, but it does entail that

there is something to objects that we, as discursive minds, are missing

and God is not. Perhaps, then, Allison’s point is that the theocentric

model is the assumption that knowledge for human beings consists in

97 Cf. B71 and a long passage in the Pölitz lectures on theology (Ak. 28: 1053-4) where Kant expresses agnosticism about whether, after death, our separated souls might be able to cognize things in themselves.

83

approximating this divine model, that for us to know some content is

to approximate to the relation that God stands in to that object. If so,

he would be right that Kant does not have a theocentric model of

knowledge, but then neither do Kant’s empiricist predecessors.

Allison anticipates this objection by arguing that Locke is committed

to a theocentric model of knowledge, but all that Allison demonstrates

is that Locke (like Kant) thinks that there is a kind of knowledge of

objects we lack and God has, something almost any theist has to

accept. Locke’s own definition of knowledge for human beings,

quoted by Allison, beings does not depend upon his conception of

divine knowledge.98

4.5 Metaphysical ‘Dual Aspect’ Readings

One prominent strand in recent scholarship on Kant’s transcendental

idealism has been the development of quite sophisticated

interpretations that try to retain the original insight that the

appearance/thing in itself distinction is not a distinction between two

different kinds of objects, while abandoning Allison-style ‘epistemic’

readings. These interpretations take the distinction to be a

metaphysical one between two different sets of properties had by one

and the same set of objects. These metaphysical ‘dual aspect

98 Allison (2004), 31-32.

84

interpretations differ in exactly how they understand the distinction

between these different sets of properties.99

Perhaps the most influential metaphysical but non-phenomenalist

interpretation of Kant’s idealism has been Langton (1998). Langton

begins by pointing out that Kant thinks we are genuinely missing out

on something in not knowing things in themselves, and this sense of

‘epistemic loss’ is incompatible with Allison’s reading. As I argued

earlier, ‘Allisonian humility’ is apparently compatible with it being

impossible that there are non-spatiotemporal objects and our forms of

intuition being the only possible such forms. This loses Kant’s sense

that we are genuinely cognitively deprived, that there is something

about the world of which we are irremediably ignorant.100

Having rejected Allison’s epistemic reading, Langton goes on to

discuss a familiar tension between two of the central doctrines of

Kant’s transcendental idealism:

(Existence) Things in themselves exist.

(Humility) We cannot know anything about things in

themselves.

Langton’s solution to this, one of the oldest problems of Kant

scholarship, is to interpret things in themselves as substances with

99 See also Allais (2004), (2006), (2007); Rosefeldt (2007) and (forthcoming); McDaniel (Ms); and Marshall (forthcoming).

100 Allison responds to Langton’s criticism in (2004), 9-11.

85

intrinsic properties, and talk of ‘phenomena’ as talk of the extrinsic

properties of those substances (things in themselves). So in general,

(1) Things in themselves are F = F is among the intrinsic

properties of substances

(2) Appearances are F = F is among the extrinsic properties of

substances

In particular, this allows Langton to interpret (Existence) and

(Humility) as:

(Existence*) Substances with intrinsic properties exist.

(Humility*) We cannot cognize the intrinsic properties of

substances.

The apparent tension between these doctrines has vanished.

Langton’s interpretation also allows her to explain why the apparent

tension between Humility and

(Non-spatiality) Things in themselves are not spatial.

is merely apparent because, on he reading (Non-spatiality) is

equivalent to:

(Non-spatiality*) Being spatial is not an intrinsic property of

substances.

This is compatible with (Humility*) because we can know it merely by

knowing that being spatial is an extrinsic property in general (thus is

not an intrinsic property had by substances), and to know this we do

86

not need to know anything about the intrinsic properties of

substances.

Langton’s explanation of how Humility is compatible with

(Affection) Things in themselves causally affect us.

requires her to slightly amend the ‘translation’ rules from above,

though, because otherwise Affection would entail that ‘affecting us’ is

an intrinsic property of substances, which she would deny. Langton’s

reconstruction of Kant’s argument for Humility rests on the premise

that the causal powers of substances by which they affect us do not

supervene on their intrinsic properties. Consequently, affect us is not

an intrinsic property of substances. Nonetheless, Langton can

reinterpret Affection as:

(Affection*) Substances causally affect us in virtue of their powers, which do not supervene on their intrinsic properties.

consistent with the spirit of her interpretation. Langton thus offers a

consistent, elegant interpretation of transcendental idealism that

solves several of the oldest and hardest problems in the interpretation

of Kant’s philosophy.

4.6 Problems for Langton’s reading

87

Much of the critical reaction to Langton (1999) has focused on her

reconstruction of Kant’s argument for Humility, but I am not going to

discuss that argument; even if Langton is wrong about how Kant

proves Humility, she may still be right about what Humility means and

thus what the appearance/thing in itself distinction means.101

4.6.1 Textual evidence

There is substantial textual evidence that Kantian appearances have

only extrinsic properties. For instance, this passage from the

Aesthetic:

everything in our cognition that belongs to intuition (with the exception, therefore, of the feeling of plea sure and displeasure and the will, which are not cognitions at all) contains nothing but mere relations, of places in one intuition (extension), alteration of places (motion), and laws in accordance with which this alteration is determined (moving forces). But what is present in the place or what it produces in the things themselves besides the alteration of place, is not given through these relations. [. . .] (B67)

This, and other passages Langton cites, support attributing to Kant

these theses:

(1)Appearances (phenomena) have only relational properties. (B67, A265/B321, A285/B341)

(2)When we conceive of an object purely intellectually we conceive of it as having intrinsic properties. (A274/B330, A277/B333)

(3)In knowing relational properties we do not know things as they are in themselves. (B67)

101 E.g. Allais (2006).

88

However, in none of these passages does Kant directly state the

stronger claim that:

(4)Phenomena are extrinsic properties of substances with intrinsic

properties.

It is clear that holds (1)-(3) and less clear that he holds (4). The

textual case for (4) is weaker, though not absent. It is presented

below, in sub-section 4.6.2.

There is a further textual problem for Langton’s interpretation,

though. In at least two passages Kant denies that we can know

relations between things in themselves:

Space represents no property at all of any things in themselves nor any relation of them each other [. . .] (A26/B42)

[. . .] the things that we intuit are not in themselves what we intuit them to be, nor are their relations so constituted in themselves as they appear to us. (A42/B59

In these passages Kant claims that space is not a relation among

things in themselves, nor are relations among objects ‘in themselves’

as they appear to us. This is hard to square with Langton’s reading.

However, in her (2011) Langton responds to these textual objections

by suggesting that the relations among things in themselves of which

Kant speaks are internal relations, relations that supervene on the

intrinsic properties of substances. This requires a slight tweak in her

definition of things in themselves and appearances:

(Things in themselves*)Things in themselves are substances with intrinsic properties, and extrinsic properties that supervene on intrinsic properties.

89

(Appearances*) Appearances are non-supervenient extrinsic properties of substances.

However, this is consistent with Langton’s overall reading because we

can now say: our ignorance of things in themselves is ignorance of

their intrinsic properties and the extrinsic properties that supervene

on them. Since causal powers (on Langton’s reading) are non-

supervenient extrinsic properties, we retain the result that things in

themselves do not affect us.

4.6.2 Phenomena substantiata

One source of resistance to Langton’s interpretation is that Kant

argues at length in the First Analogy of Experience that the category

substance can be applied to phenomena: “all appearances contain that

which persists (substance) as the object itself, and that which can

change as its mere determination, i.e., a way in which the object

exists” (A182). This would appear to contradict Langton’s assertion

that things in themselves are substances, while appearances

(phenomena) are merely properties of substances.

Langton is well aware that Kant accepts ‘phenomenal substances’

and endeavors to explain this within her picture. In doing so, she

compiles a compelling set of textual evidence for her alternative

reading of the First Analogy and the meaning of substance for

phenomena. She begins by pointing to Alexander Gottlieb

90

Baumgarten’s notion of a phaenomena substantiata, a ‘substantiated

phenomenon,’ by which Baumgarten means a property that we treat

as a substance by predicating other properties of it.102 She argues

convincingly that Kant’s fundamental notion of a substance is of being

with properties but which is not a property of anything else.103 Only

such beings, of which other things are predicated (inhere in) but

which are not predicated of (inhere in) anything else, are truly

substances. However, the properties that are predicated of

substances can also be spoken of as substances, because they

themselves have properties (which might also have properties, and so

on).104 These are substantiated phenomena.

The question is, are Kantian empirical substances genuine

substances or mere substantiated phenomena? Do the objects

subsumed under the empirical schema of substance (absolute

persistence in time) also fall under the pure category of substance

(subjects of inherence which inhere in nothing further)? If no, then

they must be predicated of some more fundamental substance, which

102 Baumgarten, Metaphysica §193 (Ak. 18:150); quoted at Langton (1998), 53.

103 In other words, the ‘pure’ concept of substance, or the ‘unschematized’ category of substance, is the concept of a being in which other beings inhere, but itself inheres in nothing further. See A147/B186, A242/B300, Refl. 5295 (Ak. 18:145).104 Langton quotes Kant’s marginal comments on Baumgarten: ‘A real subject is a substance. An accident can be a logical subject’ (Ak. 18:67).

91

drives Langton to conclude that appearances (phenomena) are

properties of substances. 105 Langton assembles an impressive array

of evidence that Kant does regard empirical ‘substances’ as

phaenomena substantiata.106 However, in context it is not clear

whether Kant has the Baumgarten notion in mind, or whether this

Latin expression means simply: phenomenal substance. So it is

unclear, textually, whether phenomena are predicated of noumena in

the Critique.

Langton does cite Kant’s claim at A525/B553 that matter is not “an

absolute subject.” But Kant there contrasts being “an absolute

subject” with being “a sensible abiding picture” which is “nothing but

intuition.” Kant’s point appears to be the familiar one (constantly

reiterated in the Antinomies section) that matter exists at least partly

in virtue of being experienced and does not have an existence

“grounded in itself.”107 This does not require him to assert that bodies

are predicated of in things in themselves.

Langton appears to concede this point108 about A525/B553, but this

is potentially more damaging to her case than she admits. In his

105 She does point out the hesitant terms in which Kant describes phenomena as substances (p. 57); see A185/B228.

106 E.g. A265/B321, A277/B333, Refl. 4421, 4422, 5294, Ak. 28:209.

107 See A491-494/B519-52.

108 Langton (2004), 60.

92

metaphysics lectures, and other texts, Kant consistently distinguishes

the inherence relation (which holds between a property and a

substance of which it is predicated) and the relation of ground to

consequence:

Etwas kann causatum alterius sein und deswegen darf es nicht der Ursache als Prädikat inhärieren. Hierin irrte Spinoza. Das Wort Subsistenz ist freilich zweideutig. Man übersetzt es durch Selbständigkeit d.h. Möglichkeit zu existieren ohne Ursache, auch durch Möglichkeit zu existieren non inhaerendo. (Ak. 28:1308)109

Kant, following Baumgarten, criticizes Spinoza’s definition of

substance as “what is in itself and conceived through itself”110

because it conflates two notions: (i) a being that is not grounded in, or

caused by, anything more fundamental, and (ii) a being that does not

inhere in anything more fundamental. The second is the correct

definition of substance, according to Kant; by conflating these two

notions, Spinoza forecloses the possibility (which Kant to be actual):

there are substances distinct from God (they are not modes of God),

all of which are grounded in God.

To bring this back to Langton, we need to distinguish two different

claims:

(i) Phenomena are extrinsic properties of substances (things in themselves). They inhere in things in themselves.

109 Ak. 8:225n, 28:562, 28:779, 28:638-9, 28:1041, 28:1104f. See also, the article on ‘Inhärenz’ in the Kant-Lexikon.

110 Ethics Id3

93

(ii) Phenomena exist (at least partly) in virtue of the extrinsic properties of substances (things in themselves). They are grounded in things in themselves.

Langton attributes (i) to Kant, but her textual case appears (ii), at

best.111 This is significant, because (ii) is far less controversial. For

instance, it is in principle acceptable to the qualified phenomenalist,

because the extrinsic properties of things in themselves include

(presumably) properties like causing us to have such and such

experience.

4.6.3 Mind-dependence

Another natural objection to Langton’s reading is that she gives short

shrift to the mind-dependence of space and objects in space, the

specifically idealist elements of ‘transcendental idealism.’112 Langton

addresses this charge, and her defense is to distinguish idealism

about space from idealism about objects in space. Consider two

different idealist theses that might be attributed to Kant:

(1) The existence of space is wholly or partially grounded in the contents of human experiences.

111 Kant does sometimes refer to things in themselves as the ‘substratum’ of appearances, but he more often uses grounding and causal terminology.112 See Bird (2000); Falkenstein (2000); and Rosefeldt (2001).

94

(2) The existence of objects in space (bodies) is wholly or partially grounded in the contents of human experiences.

Langton argues that while Kant accepts (1), he does not accept (2); in

other words, he is an idealist about space but a realist about bodies in

space. This is fully compatible with Langton’s view, for she holds that

bodies just are extrinsic properties of substances, so these bodies

would exist whether or not there were minds to perceive them

(perceive those substances as standing in those relations). She

interprets space and time as a form which our mind imposes on our

experiences of those extrinsic properties (phenomena).

Although the view Langton develops is internally coherent and

plausible and defensible, there are substantial textual problem in

attributing it to Kant. Kant repeatedly claims that the existence of

bodies in space is at least partially grounded in our experience of

them, for if we did not experience them they would not exist.113

Langton does note passages like this, and argues that her

interpretation is an attempt to make the best sense possible of as

many texts as possible; there will always be some texts that do not fit.

113 Langton notes, in particular, the apparently idealist consequences of A383 but interprets it away as “an unwise and atypical aberration.” But it is anything but atypical; Kant reiterates the point that bodies would not exist without subjects to experience them at To limit attention to passages cited already: A42/B59A383, A374n, A490-1/B518-9, A520/B492-A521/B493, A494/B522, Ak. 4:354, Refl. 5086, and Refl. 5109.

95

6. One Object or Two?

Since Karl Ameriks’ classic survey of the literature, Ameriks (1982), it

has been customary to divide interpretations of transcendental

idealism into ‘two object’ readings and ‘one object’ readings. By

contrast, I have organized this article around the distinction between

phenomenalist readings, and non-phenomenalist dual-aspect readings.

In this section I want to argue that the ‘one object’/’two object’

distinction does not track what is really at issue: (to what extent) is

Kant a phenomenalist? It does not track what is really at issue

because putative ‘one object’ interpretations can be formulated as

‘two object’ interpretations without losing what is distinctive about

them.114

6.1 Preliminary Issues

The distinction between ‘one object’ and ‘two object’ readings comes

down to the question of whether appearances, in general, are

numerically identical to things in themselves: one object readers claim

114 I argue a similar point in Stang (forthcoming): the non-identity interpretation is correct, but this does not settle the substantive interpretive questions, because identity views can be translated easily into non-identity formulations.

96

they are115, and two object readers deny this. Whether all things in

themselves are numerically identical to appearances is not at issue,

for most one object readers will admit there could be things that

never appear to us (it might be misleading to call them things ‘in

themselves’ since they never appear to us, so we never consider them

as they are ‘in themselves’).116 The qualification ‘in general’ is

necessary because some ‘two object’ readers will admit that some

appearances are also things in themselves; e.g. many ‘two object’

readers will admit that, in the case of the self, there is a single object,

a thing in itself, that appears to itself as a spatiotemporal object.117

However, the characterization of these views as ‘one object’ and

‘two object’ is unfortunate, because it is not a commitment of ‘two

object’ readings that, for each appearance, there is one and only one

thing in itself that appears as that object. Nor is the other standard

moniker, ‘one world’ versus ‘two world,’ helpful, either, for ‘world’ is a

technical term in Kant’s metaphysics and has a very specific

meaning.118 One can coherently hold a ‘non-identity’ interpretation

115 Adickes (1924), 20, 27; Allais (2004), 657; Langton (1997), 13; McDaniel (ms); Westphal (1968), 120.

116 Cf. B306, where Kant seems to admit as much.

117 On this issue, see Adams (1997), Aquila (1979) and Ameriks’ discussion of Aquila in his (1982).

118 A whole that is not a part of any greater whole and whose parts stand in real connection.

97

while denying that appearances in space and time constitute a ‘world’

at all.

In this section I will argue that the distinction between ‘identity’

and ‘non-identity’ readings is relatively trivial. I will do this, first, by

arguing that avowedly anti-phenomenalist interpretations can (and

perhaps should) be understood as non-identity views. Then, I will

argue that the phenomenalist interpretation is compatible with the

‘identity’ thesis. I will conclude that, since the interpretive idea at the

heart of these readings can be captured on either ‘identity’ or ‘non-

identity’ assumptions, the real issue is not the relatively recondite one

of whether appearances are in general numerically identical to things

in themselves, but whether the phenomenalist interpretation of Kant

is correct.

6.2 Langton and non-identity

Langton’s view can be interpreted as either an identity reading or a

non-identity reading. The difference is somewhat subtle, but it has

important consequences. On the identity version of Langton (1999),

to talk about things in themselves is to predicate intrinsic properties

of substances, while to talk about phenomena is to predicate extrinsic

properties of those very substances. On the non-identity version of

Langton (1999), phenomena are numerically identical to those

98

extrinsic properties. This would be a ion-identity reading because

substances are not identical to their properties (either extrinsic or

intrinsic). By contrast, on the identity reading, an expression for a

phenomenon refers to a substance. The difference between these

readings can be illustrated by how they give truth-conditions for the

judgment that some appearance x has property F:

(Identity) x has F = F is among the extrinsic properties of x

(Non-Identity) x has F = x, an extrinsic property of some substance

y (x), has F

While Langton initially explains her view in a way that suggests an

identity reading, she in fact opts for a non-identity reading, for good

reason. Firstly, on the identity reading Kant would have to identify

subjects of predication in empirical judgments with substances. This

is problematic because it would bring substances into the world of

space and time. For instance, if I can make a judgment about this

table, then it would be a judgment about the extrinsic properties of

this table, and this table would be a substance with intrinsic

properties (although being a table would, presumably, not be one of

them). Alternately, if we identify the table as a collection of extrinsic

properties of substances, then we can go on to predicate further

properties of the table, without having to identify the substance or

substances of which the table is ultimately predicated.

99

That the ‘identity/non-identity’ dispute is not really crucial to

Langton’s interpretation is suggested by the very terms in which she

formulates her view: “the labels ‘phenomena’ and ‘noumena’ seem to

label different entities, but really they label different classes of

properties of the same set of entities.”119 But assuming (as I think we

must) that classes of properties are entities then ‘phenomena’ and

‘noumena’ are labels for different entities — namely, different classes

of properties! Langton, of course, means that the difference between

phenomena (appearances) and noumena (things in themselves) is not

as sharp as it is on the phenomenalist view: phenomena are merely

properties of noumena. Thus, while they are numerically distinct,

there is not much metaphysical ‘room’ between them. This is

important, because, even though this is technically a non-identity or

‘two object’ view, it has a much easier time with the traditional

problems that afflict phenomenalist readings: how to square Humility

about things in themselves with the assumption that they exist and

are non-spatial? What really separates Langton from phenomenalist

readings is not the recondite question of whether phenomena and

noumena are numerically identical but how much the traditional

problems for phenomenalism apply to her view.

6.3 Phenomenalist Dual Aspect Readings

119 Langton (1998), 13.

100

Some scholars have defended what might initially seem like a

contradiction in terms: a phenomenalist ‘one object’ (identity)

interpretation of appearances and things in themselves.120 On such a

view, the appearance and the thing in itself are one and the same

object, but considered with respect to different properties: the

properties we experience the object as having, and the properties it

has. On this interpretation, Kant is qualified phenomenalist because

he holds that:

(PhenomenalismP)

The core physical properties of objects in space are grounded in the contents of our experience of them.

His attitude to:

(PhenomenalismE)

The existence of objects in space is ground partly or wholly in the contents of our experience of them.

depends upon how we read it, on this interpretation. On the one

hand, we can understand it either as the ‘de re’ claim

(PhenomenalismE*)

(x)(x is an object in space the existence of x is partly or wholly grounded in our experience of x)

120 Notably, Adickes (1924). Some readers might balk at the inclusion of Adickes among the Identity readers; see, however, Adickes (1924), 20, 27. This point is also made by Robinson (1994), 416 note 22. Also, Westphal (1968).

101

in which case Kant would reject it, because each such object in space

is also a thing in itself and, as such, does not depend for its existence

on our experience of it. On the other hand, we could understand it as

the de dicto claim

(PhenomenalismE**) The fact that there are objects in space is partly or wholly grounded in our experience of objects in space.

in which case Kant would accept it, because there being objects in

space depends upon our experiencing objects as in space.

This leads to an important exegetical point. One of the main

motivations for ‘non-identity’ interpretations are passages in which

Kant claims that appearances would not exist if there were not

subjects to experience them, e.g.:

We have therefore wanted to say that all our intuition is nothing but the representation of appearance; that the things that we intuit are not in themselves what we intuit them to be, nor are their relations so constituted in themselves as they appear to us; and that if we remove our own subject or even only the subjective constitution of the senses in general, then all the constitution, all relations of objects in space and time, indeed space and time themselves would disappear, and as appearances they cannot exist in themselves, but only in us. (A42/B59)

This might be thought to directly entail phenomenalism, for, if

appearances would not exist without subjects to experience them, but

things in themselves would, then a fortiori appearances and things in

themselves are distinct. This line of reasoning can be represented

formally as:

102

(1)For all x, if x is an appearance, then x would not exist if there were not subjects to experience it.

(2)For all x, ff x is a thing in itself, then x would exist even if there were not subjects to experience it.

(3)(x)(x is a thing in itself ~x is an appearance)121

But the Identity reader can interpret Kant’s claim “if I were to take

away the thinking subject, the whole corporeal world would have to

disappear” as meaning: without subjects to experience them,

appearances would not exist as appearances, i.e. would not appear.

In other words, she can reinterpret (1) as:

(1*) For all x, if x is an appearance, then x’s existence as an appearance (=the fact that x exists and appears) is grounded in the contents of experience.

But the conjunction of this and (2) does not entail (3); they are

compatible with the identity reading. These passages do not force the

non-identity interpretation on us.

While such an interpretation has been attractive to many readers

of Kant, it is subject to an apparent problem. If the appearance/thing

in itself distinction is a distinction between the properties objects

appear to have, and the properties they really have, it follows that for

any property we cognize of an object, that object does not really have

that property in itself. This means that, for instance, our

121 Note that this argument is valid only if ‘x would not exist if there were not subject to experience it’ is an extensional context. In Stang (forthcoming) I show that this assumption is optional.

103

representation of objects as spatial is systematically mistaken, the

very result Kant wanted to deny.

However, this objection is not nearly as damaging as is sometimes

thought. For, drawing on the distinction between perception and

universal experience from section 3, the phenomenalist identity

reader can distinguish two different ways in which a representation

can be illusory:

(1)x is perceived as F but is not F

(2)x is represented in universal experienced as F but is not F

The phenomenalist then has two options: she can either argue that (2)

is impossible, or she can argue that (2) is not a case of illusion, i.e.

that

(3)An illusion is a case in which an object x is perceived to be F but not experienced to be F.

In other words, the phenomenalist can draw a distinction between

illusion and veridical perception in terms of the content of perception

and universal experience itself. I will discuss the first strategy first.

In denying the possibility of (2), the phenomenalist is denying that,

for the range of properties that can appear in the content of

experience, it is possible for an object to lack those properties if it is

experienced as having them. The most natural ways of explaining this

is to claim that the possession of such properties by empirical objects

is grounded in our universal experience of them as possessing such

properties, i.e.

104

(4)For any experiencable property F and any object x, if x has F, then this is so in virtue of the fact that universal experience represents x as F.

Two points need to be made about this. One, it requires the

phenomenalist to alter slightly their view from my initial formulation.

The distinction between the appearance and the thing in itself is not a

distinction between the properties the object appears to have and

those it has; it is the distinction between the properties of the object,

the possession of which are grounded in experience of the object, and

those which are not. In contemporary terms, it is a distinction

between the mind-dependent and the mind-independent properties.

Secondly, the phenomenalist needs to explain what universal

experience is, and what its contents are. I explored one such

interpretation of ‘universal experience’ in section 3 above.

The alternate strategy for the phenomenalist identity reader would

be to claim that it is not a case of illusion if we experience an object as

F but that object is not F. The phenomenalist might draw a

distinction between ‘transcendental illusion’ and ‘empirical illusion’ as

follows:

Transcendental illusion. Transcendental illusions are cases in which

universal experience represents objects as having certain properties

but they lack those properties.

105

Empirical illusion. Empirical illusions are cases in which a subject

perceives an object as having a property but universal experience

does not represent it as having that property.

The phenomenalist, once again, owes us an account of what

‘universal’ experience is, and what its content is, but even without this

we can see the point of the distinction. Transcendental illusion is

opposed to cases in which we veridically experience the properties

objects really have, which, on this reading, Kant thinks is impossible.

So experience is a systematic illusion, but only in the transcendental

sense. Empirical illusion is opposed to cases in which our perceptions

cohere with one another according to the a priori laws that constitute

universal experience. In denying that the spatiality of objects is an

illusion, Kant is merely denying that it is an empirical illusion; given

that space is the form of outer sense, we know a priori that, whatever

its specific content, universal experience will represent objects as in

space. And that coherence among our perceptions is all that is needed

to ground the empirical judgment that objects are in space.

6.4 Why Identity is not the real issue

I have argued that the resolutely anti-phenomenalist reading of

Langton (1998) and the phenomenalist reading can be re-interpreted

as, respectively, a non-identity reading and an identity reading. This

106

might suggest that interpretive options are simply more complex than

has been appreciated:

Non-

Identity

Identity

Epistemic Metaphysical

Phenomen

alist

Aquila (1983), Van Cleve (1999), cf. section 3

N/A Adickes (1924), Westphal (1968)

Anti-

phenomen

alist

Alternate version of Langton

Allison (1983/2004), Bird (1962), Prauss (1974)

Langton (1998), Allais

(2006)

But the distinction between the two different versions of Langton, and

between the non-identity version of phenomenalism and the identity

version of phenomenalism is relatively trivial. It depends the question

whether appearances and things in themselves are numerically

identical, and I have argued that this question is not an important one.

So I propose that we collapse those different positions, and see the

menu of interpretive options as:

Anti-Phenomenalist

Phenomen

alist

Epistemic Metaphysical

Aquila (1983), Van Cleve (1999), Adickes (1924), Westphal (1968), cf.

Allison (1983/2004), Bird (1962), Prauss (1974)

Langton (1998), Allais (2006), alternate version of Langton

107

section 3

7. Things in themselves, noumena, and the transcendental

object

Up to this point, we have focused primarily on the nature of Kantian

appearances, and their relation to things in themselves, questions (a)

and (c) from section one. However, one of the main questions that

must be answered in any interpretation of Kant’s transcendental

idealism is, what are things in themselves? Obviously, different

interpretations will give very different answers to this question:

Phenomenalist interpretations. Perhaps the best statement of the

phenomenalist interpretation of things in themselves is given by Erich

Adickes: things in themselves are a plurality of mind-independent

centers of force.122 On this view, things in themselves are just what

we pre-theoretically took ordinary spatiotemporal objects to be:

objects that exist, and possess their core physical properties, wholly

independently of our representations of them, and which are (among)

the causal inputs to our perceptual faculties.123

Epistemic interpretations: On the epistemic reading, things in

themselves are simply objects considered independently of our

122 Adickes (1924), 14-19.

123 A variant of this thought is expressed by Ameriks (2003), 23-25.

108

distinctively spatiotemporal form of intuition. Thus, they are objects

considered as objects of a discursive cognition in general. This very

abstract thought is not the basis of any cognition, however; it is

merely a reminder that space and time are epistemic conditions,

without which we cannot cognize any object.

Metaphysical ‘dual aspect’ interpretations. On this family of

interpretations, things in themselves are objects with a given set of

properties. Different interpretations give a different answer as to

which set of properties constitute things ‘as they are in themselves.’

On Langton’s reading, for instance, things in themselves are

substances with intrinsic properties.

In this section I want to distinguish ‘things in themselves’ from

other, closely related Kantian notions: nouemena, and the

‘transcendental object.’

7.1 Phenomena and noumena

In the section ‘On the ground of the distinction of all objects into

phenomena and noumena,” which he substantially revised for the B

Edition, Kant reiterates his argument that we cannot cognize objects

beyond the bounds of possible experience, and introduces a complex

distinction between phenomena and noumena.

109

Fortunately, it is relatively clear what phenomena are:

“appearances to the extent that as objects they are thought in

accordance with the unity of the categories are called phenomena”

(A249). Earlier, in the Aesthetic, Kant had defined appearance as:

“the undetermined object of an empirical intuition” (A34/B20). I take

this to mean that appearance is the genus of which phenomenon is the

species. All objects of empirical intuition are appearances, but only

those that are “thought in accordance with the unity of the

categories” are phenomena. For instance, if I have a visual after-

image or highly disunified visual hallucination, that perception may

not represent its object as standing in cause-effect relations, or being

an alteration in an absolutely permanent substance. These would be

appearances but not phenomena. The objects of ‘universal

experience,’ as defined in section 3, are phenomena because the

categories determine the a priori conceptual form; universal

experience represents its objects under the unity of the categories.

Kant’s then introduces the concept of noumena: “if, however, I

suppose that there be things that are merely objects of the

understanding and that, nevertheless, can be given to an intuition,

although not to sensible intuition (as coram intuiti intellectuali), then

such things would be called noumena (intelligibilia)” (A249). The

concept of a noumenon, as defined here, is the concept of an object of

cognition for an intellect that is not, like ours, discursive, and thus has

110

a non-sensible form of intuition, which Kant here designates

‘intellectual intuition.’124 A sensible intuition is one that can only

intuit objects by being causally affected by them; a non-sensible

intuition is one in which the intuition of the object brings the object

into existence. Thus, the concept of a noumenon is the concept of an

object that would be cognized by an intellect whose intuition brings

its very objects into existence. Clearly, we do not cognize any

noumena, since to cognize an object for us requires intuition and our

intuition is sensible, not intellectual.

Kant then connects the concept of noumena to things in

themselves:

it also follows naturally from the concept of an appearance in general that something must correspond to it which is not in itself appearance, for appearance can be nothing for itself and outside of our kind of representation; thus, if there is not to be a constant circle, the word "appearance" must already indicate a relation to something the immediate representation of which is, to be sure, sensible, but which in itself, without this constitution of our sensibility (on which the form of our intuition is grounded), must be something, i.e., an object independent of sensibility. Now from this arises the concept of a noumenon, which, however, is not at all positive and does not signify a determinate cognition of something in general, in which I abstract from all form of sensible intuition. (A251-2)

This passage begins with the familiar point that the very concept of

appearance requires that there be something that is not appearance 124 For more on intellectual intuition in the Critique see Bxl, B145, B150, A 252/B 308–9, and A 256/B 311–12.A279/B335, B308 and Kant’s marginal notes in his own copy (E CXXX and CXXXI at A248; Ak. 23:36; E CXXXVII at A253, Ak. 23:49). The locus classicus for intellectual intuition and the closely notion of intuitive intellect [intuitiver Verstand] is the Critique of Judgment, §76-77 (Ak. 5:401-410). For critical commentary see Gram (1981), Westphal (2000), and Förster (2012), 140-143.

111

that appears. Usually Kant makes this point using the concept ‘things

in themselves.’125 However, here he claims that this idea — that it

cannot be ‘appearances all the way down’— brings with it the idea of

noumena. This is puzzling. Why must whatever it is that appear to us

as phenomena be conceived of as an objects of intellectual intuition?

Kant clarifies precisely this point in the B Edition by distinguishing

between a positive and a negative sense of ‘noumena’:

If by a noumenon we understand a thing insofar as it is not an object of our sensible intuition, because we abstract from the manner of our intuition of then this is a noumenon in the negative sense. But if we understand by that an object of a non-sensible intuition then we assume a special kind of intuition, namely intellectual intuition, which, however, is not our own, and the possibility of which we cannot understand, and this would be the noumenon in a positive sense. (B307)

Noumena in a positive sense are simply noumena as Kant originally

defined that notion in the A edition: objects of an intellectual (non-

sensible) intuition. The negative concept of noumena, however, is

simply the concept of objects that are not spatiotemporal (not objects

of our sensible intuition, namely space and time). But then it follows

that things in themselves are noumena in the negative sense,

retrospectively clarifying the passage from the A edition quoted

immediately above, where Kant seems to draw from the

Transcendental Aesthetic the conclusion that there are noumena: the

concept of appearance requires that something appears, and this

125 E.g. in the Prolegomena (Ak. 4:314-5). Cf. Bxxvi-xxvii, B306, and B307.

112

must be a negative noumena. The concept of negative noumena is a

more general concept than the concept of things in themselves

(whatever it is that appears to us as the phenomena of experience and

does not exist in virtue of being experienced) and positive noumena.

If X is the object of an intellectual intuition (a positive noumenon)

then a fortiori it not the object of our sensible intuition, so it is a

negative noumenon. This is a point about the relations among these

concepts; it holds whether or not they are possibly instantiated.

This also brings into sharp focus one of the main interpretive

questions about this passage in particular and Kant’s whole theory of

transcendental idealism in general: is Kant positively committed to

the existence of noumena (in either sense)? In a long passage, Kant

writes:

The concept of a noumenon, i.e., of a thing that is not to be thought of as an object of the senses but rather as a thing in itself (solely through a pure understanding), is not at all contradictory; for one cannot assert of sensibility that it is the only possible kind of intuition. Further, this concept is necessary in order not to extend sensible intuition to things in themselves, and thus to limit the objective validity of sensible cognition (for the other things, to which sensibility does not reach, are called noumena just in order to indicate that those cognitions cannot extend their domain to everything that the understanding thinks). In the end, however, we have no insight into the possibility of such noumena and the domain outside of the sphere of appearances is empty (for us), i.e., we have an understanding that extends farther than sensibility problematically, but no intuition, indeed not even the concept of a possible intuition, through which objects outside of the field of sensibility could be given, and about which the understanding could be employed assertorically. The concept of a noumenon is therefore merely a boundary concept, in order to limit the pretension of sensibility, and therefore only of negative use. But it is nevertheless not invented arbitrarily, but is rather connected with the limitation of sensibility, yet without being

113

able to posit anything positive outside of the domain of the latter. (A54-5/B310-11)

Kant seems to mean that the concept of noumena is merely

problematic, i.e. it serves merely to contrast our sensible mode of

knowing objects with a mode of knowing objects that would extend

beyond the limits of our sensibility. Consequently it cannot be used

‘assertorially’ to assert the existence of such objects, or to make any

determinate claims about their nature.

If this is Kant’s point, several questions arise: (i) does he mean

noumena in the positive, or in the negative, sense? (ii) if he means

negative noumena, how is this compatible with his assertion that “the

understanding, just by the fact that it accepts appearances, also

admits to the existence of things in themselves” (Prol., Ak. 4:314-5)?

This second question assumes, of course, that things in themselves

are negative noumena (on the dual aspect view: considered as they

are in themselves, objects are negative noumena, i.e. not objects of

our sensible intuition), but this connection is well established by

numerous other passages, for instance:

[. . .] our concepts of substance, force, of action, of reality, etc. are wholly independent of experience [. . .] and so in fact seem to refer to things in themselves (noumena). (Prol. §32, Ak. 4:315)

If, however, we consider both body and soul as phenomena, merely, which is not impossible, since both are objects of sense, and bear in mind that the noumenon which underlies this appearance, i.e., the outer object, as thing-in-itself, may perhaps be a simple being. (What real progress?, Ak. 20:308)

Things in themselves are negative noumena, so Kant is positively

114

committed to the existence of some negative noumena. This suggests

that in the long passage quoted above, Kant has positive noumena in

mind as the ‘boundary’ concept that cannot be used assertorically, a

hypothesis confirmed by the immediately following paragraph:

The division of objects into phaenomena and noumena, and of the world into a world of sense and a world of understanding, can therefore not be permitted at all <in a positive sense>, although concepts certainly permit of division into sensible and intellectual ones [. . .] Nevertheless the concept of a noumenon, taken problematically, remains not only admissible, but even unavoidable, as a concept setting limits to sensibility. But in that case it is not a special intelligible object for our understanding; rather an understanding to which it would belong is itself a problem, that, namely, of cognizing its object not discursively through categories but intuitively in a non-sensible intuition, the possibility of which we cannot in the least represent. (A256-7/B310-11)

Kant added the qualification <in a positive sense> in the B Edition.

Coming immediately after the previous passage (‘therefore’) this

suggests that he has the same notion of noumena in mind in both

passages, namely, positive noumena. His point in this passage is that

the concept of a positive noumena cannot be used assertorically (we

cannot assert that such objects exist, or are even possible) but is

merely serves to distinguish our sensible intuition and its objects from

a hypothetical intellectual intuition and its objects. If this is correct,

the previous paragraph, which denies any assertoric use of the

concept of noumena, is compatible with an assertoric use of negative

noumena. To answer our original question: Kant is positively

115

committed to the existence of negative noumena, but not to positive

noumena.126

7.2 The transcendental object = X

In the ‘Phenomena and noumena’ section, Kant distinguishes the

concept of a noumenon from the concept of a ‘transcendental object’

(A250). This is a reference to a notion introduced in the A version of

the Transcendental Deduction:

The pure concept of the transcendental object (which in all of our cognition is really one and the same = X) is that which in all of our empirical concepts in general can provide relation to an object, i.e. objective reality. Now this concept cannot contains any determinate intuition at all, and therefore contains nothing but that unity which must be encountered in a manifold of cognition insofar as it stands in relation to an object. (A109)127

The ‘concept of a transcendental object’ might be fruitfully thought of

as ‘the transcendental concept of an object’: the concept of ‘object’

that makes experience possible. Our mind’s synthesis of

representations into experience of objects is guided and made

possible by the idea that there is a way objects are that must be

tracked by our representations of them. This wholly abstract concept

126 For alternate interpretations of this section, see Allison (2004), 57-64, as well as Emundts (2010). Allais (2006, 146) objects against phenomenalist interpretations that Kant denies that things in themselves are positive noumena (B307), but that is not a commitment of phenomenalist readings.

127 Cf. A104.

116

of ‘a way things are’ is the concept of the transcendental object = X,

the indeterminate concept of the ‘target’ of our representational

activity. Consequently, the concept of the transcendental object must

be distinct from the concept of ‘things in themselves’ or ‘negative

noumena.’ The concept of things in themselves is the concept of the

(unknowable by us) objects (or aspects of objects) that appear to us

the 3D world of space and time. They are the grounds of phenomena,

while the transcendental object is the very abstract idea of those

objects in space and time as the targets of our cognitive activity.

Another way to appreciate this distinction is to consider the

difference in why these notions of object (noumena, transcendental

object) are unknowable by us. We cannot cognize things in

themselves because cognition requires intuition, and our intuition

only ever presents phenomena, not things in themselves. We cannot

cognize the transcendental object because the transcendental object

is a purely schematic, general idea of empirical objectivity. Whenever

we cognize a determinate empirical object we are cognitively

deploying the transcendental concept of an object in general, but we

are not coming to know anything about the object of that concept as

such.

This is Kant’s point in ‘phenomena and noumena’ when he writes:

This transcendental object cannot even be separated from the sensible data, for then nothing would remain through which it would be thought. It is therefore no object of cognition in itself, but only the representation of appearances under the concept of an object in

117

general, which is determinable through the manifold of those appearances. (A250-1)

The (negative) concept of a noumenon is the concept of an object that

is not an object of our sensible spatiotemporal intuition. But the

transcendental object makes no sense in abstraction from intuition,

because it is merely the abstract concept that the unity of our

intuitions must have in order to constitute experience of an object.128

Conclusion

This article has traced the meaning of transcendental idealism,

sometimes referred to as ‘critical’ or ‘formal’ idealism, through the

text of the Critique of Pure Reason and various interpretive

controversies. I have argued that the main question dividing different

interpretations is whether Kant is a phenomenalist about object in

space and time and, if so, in what sense. The phenomenalist

interpretation of Kant, dominant among Kant’s immediate

predecessors and later German idealists, was challenged in twentieth

century Anglophone scholarship by, among others, Graham Bird,

Gerold Prauss, and Henry Allison. Some later scholars have retained

a central idea of these scholars’ reading —that the appearance/thing

in itself distinction is a distinction between distinct aspects of objects,

128 Cf. Allison’s classic paper on the transcendental object, Allison (1968).

118

not distinct kinds of objects — while jettisoning the purely

epistemological interpretation of Kant’s idealism, made famous by

Allison. I have also argued that one way of characterizing these

debates about transcendental idealism — as between those who hold

that appearances are identical to things in themselves, and those who

deny this — is unhelpful. What is really at stake in debates about the

meaning and philosophical legacy of Kantian transcendental idealism

is the question raised and debated by Kant’s first readers: is this a

form of phenomenalism?

119