. 2
( 9)


own states.± Kant concedes that there can be representations of which
the creature who has them can have no cognition:

I would not even be able to know that I have them [sense data], and they would
therefore be nothing for me, as a knowing being, at all. They might still exist in
me (if I imagine myself to be an animal) a being unconscious of my own
existence carrying on their play in an orderly fashion as representations
connected according to an empirical law of association, exercising in¬‚uence
upon feeling and desire, and so always displaying regularity without my thereby
acquiring the least knowledge of anything, not even of these my own
states. (May , ±·, Ak. ©, p. µ)

Kant™s willingness to allow for association even if a creature has no
concepts of objects seems at ¬rst to con¬‚ict with his account of the
threefold synthesis.± For it seems to provide a counterexample to the
twin theses (±) that one needs the capacity for conceptual recognition in
order to be able to distinguish an object from the way it appears to one
and () that one then also needs the ability to distinguish an object from
the way it appears to one in order to have some determinate basis for
associations. But, in fact, the existence of associative capacities in ani-
mals that do not have concepts is not obviously inconsistent with the
account of association in the A-Deduction. One can consistently argue
that association requires the kinds of regularities that one can only make
sense of by using concepts of objects, while also arguing that association
is possible even for creatures that cannot themselves use concepts. This
is also the key to understanding how it is possible to accommodate the
representations of animals in the possibility of self-consciousness.
An animal representation is a representation of an object from a
certain spatio-temporal standpoint. The recurrence of certain perspec-
tival presentations of a spatio-temporal object for that animal is the basis
for the associations that Kant regards as key to animal consciousness.
Now the animal cannot itself become conscious of the various perspec-
tival presentations of the object with which it is presented. At least it
cannot regard those presentations as its own distinctive take on the
object. For the animal has neither self-consciousness, nor a bona ¬de
concept of an object that can be identi¬ed and re-identi¬ed in di¬erent
circumstances. But the very possibility of thinking of an animal as
associating di¬erent spatio-temporal perspectives on an object is some-
thing that must be intelligible. Animals do not themselves distinguish
Introducing apperception
objects from the way those objects immediately present themselves to
those animals. We can make sense of the distinction between the object
and the way it appears to an observer only by thinking of the observer as
if it were the kind of creature for whom the object could be something.
This means we must think of the creature as if the creature were
conscious of its own distinctive point of view, and this means that we
must think of the creature as if it were self-conscious. To do this, we
must, as selfconscious beings, put ourselves in the vantage point of the
non-selfconscious being in question. In this way, we are able to treat its
non-selfconscious representations as if they were self-conscious repre-
sentations and thus regard them as potential candidates for selfcon-
Animals have no self-knowledge. In this sense, their representations
are nothing for them. While representations in such non-selfconscious
animals cannot become self-conscious to the animals that have them, we
can only attribute representations to animals based on our ability to
imagine what it would be like for us to represent the world in the way
animals do. This is why, in discussing unconscious representations,
Kant demands that I, who am a self-conscious being, imagine myself to
be a being unconscious of its existence. Thus, even though representa-
tions may not be candidates for self-consciousness by the animal that has
them, they are intelligible to us only in virtue of the fact that we think of
them as representations that we might have ascribed to ourselves had
our circumstances been quite di¬erent. For we only think of a represen-
tation as representation to the extent that we think of it as an expression
of a sentient point of view. And to think of the representation as an
expression of a sentient point of view is already to think of the represen-
tation as something that a being could, in principle, think of as its own
were its cognitive capacities like our own.°
In arguing that all representations are potential candidates for self-
consciousness at  ±±·n, Kant moves from the assumption that a repre-
sentation is unknowable to the conclusion that it does not exist at all. At
 ±° the point is put in terms of objects of representation, that is, that
which appears to us: ˜˜without the relation to at least a possible con-
sciousness, appearance would never be an object of cognition for us and
therefore nothing for us, and since it has no objective reality in itself, and
only in cognition, it would be everywhere nothing.™™ Prima facie, there is
an important distinction to be drawn between the conditions governing
our ability to ascribe content to representations and the conditions
governing the very existence of representational content. Kant argues
 Kant and the demands of self-consciousness
here that representations have no existence that is independent of the
capacity of creatures like us to recognize them. Representation is con-
stituted by its function in understanding how sentient creatures react to
their environment. As such, it is exhausted by the role it plays in
understanding persons, animals, and certain sophisticated automata
(automata spiritualia). We can regard the premise that representation is
exhausted by its cognitive role as analytic to the notion of representa-
tion. However, the premise is not simply a verbal stipulation, but a
substantive claim about the nature of representations. It can thus
equally be regarded as synthetic and even synthetic a priori.

 °    ®   © ®  ¬   ®   ®   ® ¤ ¦µ ®  ©  ®  ¬  ¬ 
Kant insists that representations must make a discernible di¬erence if
they are to count as distinctive representations. In fact, he argues that
representations ˜˜can represent something only in so far as they belong
with all others [actually, all other consciousness] to one consciousness,
and therefore must at least be capable of being so connected™™ ( ±±).
Representations represent in virtue of the di¬erent contents that they
have. They can therefore only be regarded as representational contents
to the extent that they can be distinguished from one another. But, in
order to count as representational contents, they must also be logically
distinguishable from their bearer even if they are not, as in the case of a
pain sensation, logically distinguishable from their object. To be logi-
cally distinguishable from their bearer, representations must be think-
able as having a subject. As representational contents, they are thus
distinguishable from one another only insofar as they can be represen-
ted as di¬erent possible representations of a subject. They must, how-
ever, also be represented as such by a subject because it is only through
the possibility of ¬rst-person access that we understand what makes a
representation the qualitative experience that it is. This leads Kant to
conclude: ˜˜The abiding and unchanging ˜I™ (pure apperception) forms
the correlate of all our representations in so far as it is to be at all possible
that we should become conscious of them™™ ( ±).
In order to be able to consider any content experienced by us
as a representational content, we must be able to recognize that content
as a representational content. But we can only recognize any-
thing as a representational content if we can compare and contrast it
with other representational contents in one consciousness. Contrastive
consciousness involves a representation of many at least numerically
Introducing apperception
di¬erent representations. If those representations are to be represented
in relation to each other, one must be able to represent oneself as the
numerically same subject which is representing them in relation to each
other. From this line of thought, Kant infers that the possibility of
representing the self as numerically identical through its di¬erent poss-
ible representations is necessary to the very possibility of representa-
We are conscious a priori of the pervasive identity of our self in respect to all
representations which can ever belong to our knowledge as a necessary condi-
tion of the possibility of all representations (because these only represent
something in me through belonging with all other (mit allem anderen) [conscious-
ness] to one consciousness, therefore they must at least be connectable there-
in). ( ±±)
The content of a representation must be a content which can belong to
the interconnected representations of some self-conscious thinker. The
content must be a candidate for self-consciousness by some possible
self-conscious being in order for there to be grounds for the attribution
of that representation to a representer. We cannot however represent
the identity and di¬erences of these di¬erent representations in relation
to each other without applying concepts to those representations. A
concept just is a representation of an object by means of a feature which
may be shared by di¬erent objects (including representations them-
selves). These concepts must themselves have contents which are deter-
mined by their relations of identity and di¬erence to each other. Thus,
we can only make sense of the notion of a representation insofar as we
hold that representation to be conceptualizable within the conceptual
framework of a being capable of articulating a set of concepts which it is
capable of self-ascribing. The further point to note is that the systematic
di¬erences between representational contents can be understood as
di¬erences in the functional roles that representations play in judgment
and inference. It is here that the categories enter the picture.

¬§©¬ ¦µ®©® ¦ µ§ ®¤ §©
The categories or pure concepts of the understanding whose legitimacy
Kant wishes to defend are supposed to have their ultimate source in the
basic logical constants or logical functions of thought. The Metaphysical
Deduction claims to be able to derive the categories from the most basic
logical functions involved in judgment. This list of categories and basic
logical functions is supposed to be complete:
 Kant and the demands of self-consciousness
This division is systematically generated from a common principle, namely
from the faculty of judgment (which is as much as the faculty of thinking) and
has not come to be rhapsodically from a search for pure concepts relying on
good luck, of whose completeness one could never be certain, since it is only
inferred by induction without considering that in that case one could never
realize why these concepts and not others belong to pure understanding. (
°“±/ ±°“±°·)

Although Kant chastises Aristotle for his rhapsodic approach to a
doctrine of categories, it is di¬cult to identify an argument in the
Metaphysical Deduction or elsewhere that the table of logical functions
or of categories is complete. Kant seems, however, to have thought that
an analysis of the nature of judgment would force one to assume a
certain set of most basic syntactical forms of judgment. To such syntacti-
cal forms of judgment there would in turn be a corresponding set of
most basic extra syntactic constraints on the way judgments can be
made concerning objects.±
Kant lists twelve categories in four groups, (±) categories of quantity:
unity, plurality, totality, () categories of quality: reality, negation, limi-
tation, () categories of relation: inherence and subsistence, causality
and dependence, interaction, () categories of modality: possibility,
actuality, and necessity ( °/ ±°). These categories are supposed to
arise from the di¬erent ways in which intuition can be connected
together in a unity, just as the twelve basic forms of judgment that Kant
attempts to abstract from term logic are supposed to arise from the
di¬erent fundamental ways in which concepts can be connected to-
gether in the unity of a judgment ( ·/ ±°µ). The general idea is that
categories constitute the di¬erent fundamental kinds of objects that can
be thought by the di¬erent fundamental functions involved in di¬erent
kinds of judging. These forms of judgment are (±) quantity of judgment:
universal, particular, individual, () quality: assertive, negative, in¬nite,
() relation: categorical, hypothetical, disjunctive, () modality: problem-
atic, assertoric, apodictic. Even if Kant had a completeness proof for the
logical constants in his term logic, the connection between the catego-
ries and these logical functions is quite loose. Questions would thus
remain about the completeness of his list of categories.
While categories have a logical meaning that consists in the uni¬ca-
tion of representations in judgment even without sensible conditions,
these logical forms are not yet concepts of objects ( ±·/ ±). They
only become full-¬‚edged concepts of objects when they are given an
extra syntactical content from our experience. For until then the objects
Introducing apperception
that these forms represent are indistinguishable from those forms of
thought themselves. Once logical forms have an object that is distin-
guishable from the forms of thought, they become categories ( ·/
±°µ). A concept a priori that did not have a content somehow dependent
on experience and that did not relate to experience at all would be ˜˜only
the logical form for a concept, but not the concept itself through which
something would be thought™™ ( µ). While we can have concepts of
objects that do not or could not exist in our experience, even such
concepts are formed from concepts that do have a content that is
somehow linked to experience. This content must be at least a condition
for the possibility of experience ( µ“). Such concepts need not be
acquired from experience, but, in the helpful jargon of Kant™s reply to
Eberhard, they must at least be originally acquired in response to
experience and the spatial and temporal structure which makes our
experience what it is. This is what gives such concepts their ˜˜objective
reality™™ and such objective reality is required if our use of such concepts
is to be legitimate.
Now Kant does not seem initially to be willing to extend functional
role to all mental states, but maintains instead that functions consist in
ways of subsuming representations under a concept. He notes that
˜˜concepts are based on the spontaneity of thought, just as sensible
intuitions depend on the receptivity of impressions™™ ( / ). In the
same context, Kant explicitly contrasts the dependence of concepts on
functions with the way in which ˜˜all intuitions as sensible, depend on
a¬ections.™™ The functions involved in our ability to use concepts are
limited to the active taking of things as thus and such to be found
paradigmatically in judgment and inference.
But the initial picture of a sharp distinction between spontaneity and
receptivity, and function and a¬ection soon breaks down. For it turns
out that our experience is inherently concept-laden, so that even when
we are not making judgments, what we sensibly experience is already
conceptualized. While the Deduction begins with the claim that ˜˜intu-
ition does not require the functions of thought in any way™™ and poses the
problem of how categories can apply to objects of intuition given this
independence of intuition from the functions of thought ( ±/ ±), the
argument of the Deduction in favor of the legitimacy of our possession
of categories, that is, what Kant calls pure concepts of the understand-
ing, actually goes through only if Kant can establish the quite di¬erent
claim that intuition depends for its unity, that is, for its ability to
represent determinate objects, on the functions of the understanding:
° Kant and the demands of self-consciousness
˜˜The same function that gives unity to the representations in a judgment,
gives also the mere synthesis of di¬erent representations in one intuition
unity, which expressed generally, is called the pure concept of the
understanding™™ ( ·/ ±°µ).
What we directly experience or intuit has a determinate object, only
insofar as it is already conceptualized by us and thus subject to the same
functions of unity that underlie di¬erent conceptual roles. Indeed, Kant
maintains that the relation of di¬erent contents of cognition to an object
˜˜is nothing but the necessary unity of consciousness, hence also of the
synthesis of the manifold, to connect it in one representation through a
common function :gemeinschaftliche Funktion9 of the mind™™ ( ±°).
Cognition relates to an object insofar as it involves a necessary unity of
consciousness. The necessary unity of consciousness involved in cogni-
tion of an object is a way of representing things that is independent of
the psychological state one happens to be in, because it must be the
same for all of us at all times and places. This way of representing things
is provided by a function of the mind that is the same for all of us at all
Kant must insist that inner perception (introspection) involves sensi-
bility, indeed, inner a¬ection, but also that it has an active and sponta-
neous dimension to it that re¬‚ects its dependence on the functions of
thought that underlie the identity of self-consciousness. And, if Kant is
to successfully argue that the content even of intuitions is dependent on
the kind of identity of function to be found in concepts, but which
ultimately depends on our capacity for representing our numerical
identity in di¬erent experiences, he must argue that even the representa-
tional content involved in intuition is a function of the judgmental and
inferential role which that content plays in judgments and inferences
using concepts. Thus, even though the Deduction starts out by indica-
ting that objects of intuition might be completely independent of the
functions of the understanding, the conclusion of the Deduction must be
that we have no way of understanding what it would be for an object of
intuition to be an object and yet not subject to concepts.

 °    ®   © ®  ¬   ®   ®   ® ¤ ¦µ ®  ©  ®  ¬  ¬ 
So far I have treated Kant™s account of representational content as an
expression of its dependence on the functional role that such content
plays in inference and judgment. In this respect, my interpretation has
obvious similarities to Wilfried Sellars™s functional interpretation of
Introducing apperception
thought, apperception, and the subject of apperception. Appealing
apparently to the Kantian idea that ˜˜concepts are based on functions™™
( / ), and that thoughts are constituted by relations between
concepts, Sellars argues that Kant allows for no other characterization
of thought processes than in purely functional terms. He also insists that
Kant™s transcendental self is constituted by functional relations. For
Sellars, the transcendental self is an epistemic principle to which any
true thought must conform. The principle in question is that an I thinks
the thought of a temporal system of states of a¬airs and any actual state
of a¬airs belongs to this temporal system of states of a¬airs. Sellars
notes that the thought of this temporal system involves the thought of a
complex; as such it is synthetic, indeed it is the synthetic unity of
apperception. But it entails an identity of the I that thinks one content
with the I that thinks another content in the system, and thus it entails
what Kant in the B-Deduction calls the analytic unity of apperception.
Now this identity of the I is something that is characterized purely in
terms of its functional role; as such, in principle, it could be embodied in
di¬erent kinds of real bearers. Since we can characterize the transcen-
dental self only functionally, we cannot infer anything about the speci¬c
nature of the transcendent self or ultimate bearer of thought, and thus
we cannot claim that it is a material entity.µ
Because the transcendental self is understood in purely functional
terms, knowledge of that self does not presuppose matter of factual
knowledge of the self as does our access to the transcendent or noumenal
self, that is, the self as an object of understanding alone. Somewhat
surprisingly, Sellars maintains that the states of the empirical self, that is,
the states of self as it is knowable from experience, are also characterized
in purely functional terms. The claim that what we know of our selves
empirically is known in purely functional terms has a more tenuous
connection to the Kantian text.
Sellars acknowledges that Kant takes the states of the empirical self to
be ones which are passive, states that the self is in some sense caused to
have. He even thinks that Kant is legitimate in thinking of some states,
such as perceptions, as passive in that they can be taken to be caused by
the states of material substance.· Although Sellars does not do so, he
might take such states as functional in the sense that they causally
depend on certain inputs. But this would be to use the term ˜˜functional™™
in a new and di¬erent sense than the one that has a basis in the Kantian
text. Patricia Kitcher thinks of Kant as a functionalist in this causal
sense, which she links with the Kantian idea of synthesis. This has two
 Kant and the demands of self-consciousness
rather paradoxical consequences. Kitcher maintains that judgments
have representational character only on the basis of their dependence
on intuitions. In other words, as Kitcher interprets Kant, judgments
have a functional role in representation only in virtue of their depend-
ence on intuition. And synthesis, which Kant insists is a spontaneous
activity of the self, becomes a function of the causal dependence of the
self on stimuli.
Kitcher reconstructs Kant™s position along the lines of a functionalist
cognitive psychology in which personal self-consciousness is a construct
out of impersonal activities of information processing. She insists that an
understanding of the self as an agency for connecting representations
can only be made intelligible by collapsing the transcendental and the
transcendent self into the empirical self, where the transcendental self is
the self as an enabling condition of experience, the transcendent self is
the self as it exists independently of any experience, and the empirical
self is the self that we experience through the changes in our representa-
tions. Kitcher denies the existence of any subject underlying mental
states and giving them unity, but she insists on the reality of mental
states and their unity, and tries to give Kant™s epistemological doctrines
an explicitly psychological interpretation wherever possible.° Although
Kitcher does not emphasize a priori constraints on representation, she
tends to regard such constraints as expressions of innate dispositions of
the mind.
For Kitcher, the self does not determine the content of representa-
tions. Instead the self arises out of that content. The self is nothing but a
system of representations which is interconnected by the nature of its
content, a quasi-causal relation between representational states.± Des-
pite her emphasis on Kant™s response to Hume™s views on personal
identity, Kitcher ascribes to Kant a variant of the bundle theory in
which the Humean constraint of existential independence is relaxed.
Where Hume maintained that all representations are ultimately com-
pletely disconnected from one another, Kitcher rightly has Kant insist
that there are connections between representations. But, since such
connections have no intrinsic connection to a self for Kitcher, she faces a
problem that besets Hume™s conception of the self. If the self is a
construct out of a collection of representations, why then are my
representations in fact mine, that is, on what basis do we think that a
given set of representations belongs to the same self to begin with?
The status of mental unity in Sellars™s functionalist interpretation of
the self is somewhat more di¬cult to assess. Sellars denies that persons
Introducing apperception
are logical constructs out of mental or physical events. He thinks that
persons have a kind of unity to them that Kant would attribute to the
spontaneous activities of the self. Sellars rightly notes that, for Kant,
perception is not purely passive, and that we also engage in other
activities of thought that are more properly thought of as active or
˜˜spontaneous™™ in Kant™s vocabulary. I think a sympathetic reading of
Sellars would take him to be acknowledging that, whereas ˜˜thought
taken by itself is merely logical function, hence pure spontaneity of
connection of the manifold of a merely possible intuition™™ ( ), inner
experience, and hence an understanding of the whole panoply of di¬er-
ent human representational states, is ˜˜no more mere spontaneity of
thought, but also receptivity of intuition, that is, thought of myself
applied to empirical intuition of that same subject™™ ( °). In other
words, it is through the concept-ladenness of sensible experience that
the representational states belonging to such experience can themselves
be understood in terms of the functional roles that properly belong to
concepts and, more speci¬cally, to logical concepts. Logical concepts
are basic because they spell out the most basic forms of conceptual role
in judgment and inference.
For Sellars, the passivity or a¬ectedness of the mind displays itself in
the fact that the states of the empirical self belong to a deterministic
system of events that cannot be understood independently of interacting
material substances. The activity of the mind is displayed in its ability
to reason and understand, capacities that are tied up with pure apper-
ception. Sellars notes that Kant™s theoretical philosophy does not pro-
vide a compelling reason for assuming that the mind should be regarded
as anything but relatively spontaneous, that is, as able to have a take on
objects, but one which is activated by causes outside of it. However,
Sellars does note that Kant takes his notion of the autonomy of practical
reason to require a form of spontaneity that goes beyond the relative
spontaneity required by theoretical reason. And, indeed, the passage
that I have quoted from the Paralogisms concludes with the remark that
in the moral law ˜˜a spontaneity would be found through which our
reality would be determinable without requiring the conditions of em-
pirical intuition™™ ( °“±).
I have already indicated a good deal of sympathy with Sellars™s
interpretation of apperception and its transcendental subject. Here are
some caveats. While Sellars acknowledges the primacy of the transcen-
dental self in the order of knowledge and in practical deliberation about
how we are to live our lives, and even insists that the self so understood
 Kant and the demands of self-consciousness
plays a constitutive role in the manifest image of everyday experience,
he also wishes to argue that the transcendental self is, in a certain sense,
eliminable in the more fundamental scienti¬c image of nature. While
the thinking self is not simply a congery of physical events, it has no
existence that is independent of certain con¬gurations of physical
events. As such, the self is a form of necessary illusion generated by the
¬rst-person point of view. We can dissolve the illusion by shifting to the
objective point of view belonging to the scienti¬c image of human
beings. It would be unfair to say that, since the transcendental or
thinking self is not a part of the objective point of view for Sellars, it
simply drops out of any account of objectivity. But it is fair to say that
Sellars does not really develop the sense in which one can regard the
transcendental self as a basic epistemic condition of experience and it is
unclear how the transcendental self can function as a basic epistemic
condition while also having no fundamental ontological role to play in
In contrast to the Sellarsian account, I wish to show how Kant carries
out the task of developing objective constraints on experience from the
relation of di¬erent representations to the transcendental or impersonal
representation of self that we have in self-consciousness. In the next
chapter, I take up Kant™s argument to the necessary representation of
numerical identity from the need to be able to associate representations
in order to be able to have any experience at all. First I show how one
can argue with Kant that any discriminatory awareness involves the
capacity to connect di¬erent representations in time. I then develop his
argument that we can only make sense of the capacity to link di¬erent
representations in time by appeal to our ability to associate items in
experience. The key argument is, however, based on the claim that we
can only make sense of this capacity for association by appeal to the
concept of objects that are distinct from our immediate experiences.
Since Kant maintains that the concept of an object requires the possibil-
ity of linking together what we represent from the impersonal stand-
point of transcendental self-consciousness, this leads Kant to argue that
such self-consciousness is a necessary condition for any coherent experi-
ence at all. This, then, is the ultimate basis of Kant™s claim that
apperception can prove its identity with respect to di¬erent representa-
tions and that such representations are therefore subject to functions of
° 

Concepts, laws, and the recognition of objects

I claimed in chapter two that Kant™s notion of transcendental appercep-
tion involves a necessary representation of the numerical identity of the
self. I also claimed that transcendental apperception is a capacity that
Kant links to the conditions under which we could recognize any object
of a representation. I want now to turn to Kant™s argument to the
conclusion that a necessary representation of numerical identity is
required if we are to be able to recognize objects that are distinct from
our representations. Kant not only argues that transcendental apper-
ception and recognition of objects go together, he also argues that
recognition of objects must be possible if we are to make sense of even
the most minimal kind of discriminatory awareness. In this way, the
argument becomes crucial not only to a defense of objectivity, but also
to an understanding of the dependence of subjective on objective
In the ¬rst section of this chapter, I take up Kant™s so-called threefold
synthesis of apprehension, reproduction, and recognition. Here, Kant
¬rst argues that any discriminatory awareness of what we experience
involves the capacity to connect di¬erent temporal perspectives to-
gether. He then argues that this requires the ability to associate items
that we experience at one time with items that we experience at another
time. Such associations in turn require the existence of objects that are
distinct from what we experience from the vantage-point of any given
temporal perspective.
In the next section, I take up Kant™s claim that we can only make
sense of objects that underwrite associations if we think of such objects
as represented by us in the standpoint-independent way that is made
possible by our ability to use concepts. Our ability to use concepts, as
Kant understands them, depends on our capacity to represent things in
the same way from any standpoint that is intelligible to us regardless of
what that standpoint might be. I argue that Kant regards concepts as
 Kant and the demands of self-consciousness
representations that are underwritten by a normative commitment to
laws linking the properties represented by those concepts. It is these laws
that then underwrite our capacity to associate items in experience.
Kant argues that we can only distinguish objects in our experience
that underwrite associations to the extent that representations are linked
together by necessary connections provided by laws. These laws are
then the basis for our knowledge of objects. But we can only know them
because they derive their lawlikeness from our ability to represent them
as expressions of our numerical identity in any arbitrary situation. From
this, Kant concludes that cognition is nothing but the uni¬cation of
di¬erent representations in the impersonal point of view of transcenden-
tal apperception.
Perhaps the key claim in Kant™s argument is that a necessary con-
sciousness of one™s numerical identity as a self goes together with our
concept of an object of experience. This leads him to claim that we must
be able to represent ourselves as numerically identical if our experiences
are to have the kind of associability in virtue of which they may be said
to belong to one experience. Kant draws interesting and far-reaching
conclusions from his claim that experience must be conceptualizable if it
is to be an experience of objects at all. Nature must be such that we are
necessarily able to associate our di¬erent experiences of nature guided
by concepts of the objects that we experience. And this means that
nature must be regarded as uniform with respect to the concepts that we
have, where this uniformity can only mean that we are able to provide a
systematic description of nature in terms of the laws that our concepts
purport to express.

¦     °°    ®  ©  ®       §® © ©  ®
Kant distinguishes between a synthesis of apprehension, of reproduc-
tion, and of recognition, corresponding to a capacity to distinguish
di¬erent contents of consciousness, to associate di¬erent contents of
consciousness, and to recognize di¬erent contents of consciousness by
means of concepts. Ultimately, Kant wishes to argue that the categories
are the a priori concepts required for the necessary connections re-
quired by the synthesis of recognition. Such recognition is itself an
enabling condition for the apprehension and reproduction of items in
experience. In this way, he hopes to show that categories must function
as a priori constraints on recognition if we are to be able to apprehend
items at all.
Concepts, laws, and the recognition of objects
Now, in arguing for the need for a synthesis of apprehension, Kant
insists that what we experience may be inherently complex, but it is only
characterizable as complex, indeed as what it is, insofar as we are able to
distinguish ˜˜time in the successive sequence of impressions™™ ( ). To
distinguish di¬erent times, and hence also the contents of di¬erent
times, we must be able to connect the contents of our di¬erent experien-
ces together in time. In this way, we come to grasp time as a whole of
time, and also space as a whole of space. For intuition, through which
we immediately experience items in time and space, ˜˜presents a mani-
fold, but cannot give rise to this manifold as such and indeed in one
representation without a thereby occurring synthesis™™ ( ). The synthesis
of apprehension is concerned with the basic preconceptual level at
which we distinguish temporal perspectives and di¬erent items in per-
ceptual experience.
It has sometimes been suggested that the synthesis of apprehension
already involves transcendental apperception, and that no conceptual
factor is required in order to account for the possibility of self-conscious-
ness of the anticipations and retentions that go into the kind of discrimi-
nation involved in apprehension.± But this is not the way Kant himself
argues. Kant argues that transcendental self-consciousness is required
for conceptual recognition. To be able to recognize things by means of
concepts, we must be able to represent ourselves as having the numeri-
cally same point of view in di¬erent experiences. This necessary repre-
sentation of numerical identity is a transcendental consciousness of self,
that is, it is a representation of self as an enabling condition of experi-
ence. Our ability to apprehend and reproduce what we experience is
much less directly tied to transcendental self-consciousness than is our
capacity to recognize objects by means of concepts. We can only
apprehend and reproduce items in our experience because we are able
to experience objects of the kind that make the associative processes
involved in reproduction possible. But we can only make sense of the
existence of such objects by means of concepts. And these concepts, in
turn, depend for their existence on our capacity to represent ourselves as
having the same point of view in di¬erent experiences.
Kant insists that all apprehension of items at a time requires processes
of association, that is, a synthesis of reproduction, through which im-
mediately past experience is retained in the present ( ±°), and this
synthesis of reproduction ultimately turns out to presuppose the syn-
thesis of recognition, through which we recognize objects that are
independent of our momentary experiences. Finally, such recognition of
 Kant and the demands of self-consciousness
objects involves a kind of necessity that depends on transcendental
apperception. Thus, the syntheses of apprehension, reproduction, and
recognition are distinct, but mutually dependent processes of coming to
terms with perceptual information that are supposed to be guided by the
possibility of self-consciousness with respect to the experiences that
involve them.
If even the states of our immediate past are not available to us at any
given moment, then we will fail to distinguish anything from anything
else, for we will have already lost the awareness of the item from which
the second item is to be distinguished. On the other hand, such items
need to be distinguished from each other through apprehension and
recognized as identical or distinct through recognition. And Kant
claims that this requires transcendental self-consciousness.
Kant follows Locke, Hume, Tetens, and many other empiricists in
deriving our ability to connect a present state of consciousness with
some past state of consciousness from a habit of associating the one state
with the other state. Reproduction thus depends on empirical patterns
of association. Like Locke, Berkeley, and Hume, Kant assumes that the
items that we perceive and associate are appearances, that is, things as
they appear to the mind. But Kant rejects the British empiricist project
of understanding the unity of things as bundles of representations. Even
Locke, who is committed to the existence of physical objects, accounts
for our knowledge of such objects in terms of the bundling of sensations:
The mind being, as I have declared, furnished with a great number of the
simple ideas, conveyed in by the senses as they are found in external things, or
by re¬‚ection on its own operations, takes notice also that a certain number of
these simple ideas go constantly together; which being presumed to belong to
one thing, and words being suited to common apprehension, and made use of
for quick dispatch, are called, so united in one subject, by one name; which, by
inadvertency, we are apt afterward to talk of and consider as one simple idea,
which indeed is a complication of many ideas together; because, as I have said,
not imagining how these simple ideas can subsist by themselves, we accustom
ourselves to suppose some substratum wherein they do subsist, and from which
they do result, which therefore we call substance. (Essay Concerning Human Under-
standing, Book II, chapter xxiii, p. ±)

Against the associationist bundle theory, Kant insists that association
would be completely arbitrary if there were no regularities in the world
underlying the co-occurrence of our experiences. Such regularities
explain why we associate one set of representations with another set of
representations. A certain amount of regularity must be assumed in
Concepts, laws, and the recognition of objects
experience in order for us to come to associate di¬erent items of
experience. Although things remain relatively constant in experience:
If cinnabar were ¬rst red, then black, ¬rst light, then heavy, a human being
were to be changed ¬rst into this, then into that animal form, if on the longest
day the land were to be covered ¬rst with fruit, then with ice and snow, then my
empirical imagination would not even have the opportunity to think of heavy
cinnabar when representing a red color. ( ±°±)

The kind of regularity required is complex. For, even if objects persist
over time, they are also subject to change. Cinnabar turns from red to
black when oxidized, and human beings take on di¬erent animal forms
when they dress up as animals. Associations are only of interest in
grasping regularities in experience if they are based on some signi¬cant
shared feature of the items associated. Everything is similar to every-
thing else in some sense or other. So there must be some basis for picking
out one set of similarities as relevant to a set of experiences as opposed to
another set. It might seem that relevance could be randomly deter-
mined. But a random relation between cognitive contents would be ˜˜a
blind play of representations, less even than a dream™™ ( ±±). Even
dreams involve rather sophisticated forms of regularity that would be
absent from a random collection of mental states.
Kant does not argue for his key and quite astounding claim that to
account for the associability of our perceptions we must appeal to a priori
connections between what we perceive ( ±°±). The more plausible view
is that we directly perceive objects the co-occurrence of which accounts
for our dispositions to associate certain things with certain other things.
If we directly perceive objects that exist independently of us, then the
recurrence of those objects can be invoked to explain our ability to
associate items in our experience in non-arbitrary ways. Thus some
form of direct realist theory of perception would be the more obvious
solution to the problem of association than an appeal to a priori
Fortunately, Kant™s account of the synthesis of recognition attempts
to show that a direct realist theory of perception presupposes a priori
connections between the items that one experiences. His theory of
recognition thus promises to supply the rationale for an appeal to a
priori synthesis that is missing from his account of association. Kant
does not deny the claim that we directly perceive objects such as tables
and chairs that provide the basis for the associations in terms of which
we connect diverse experiences. However, the problem with such
µ° Kant and the demands of self-consciousness
perceptual objects is that they are not the sort of things that are subject
to laws. When one bases associations on perceptual objects, one bases
the process of association on objects that are themselves, at best, only the
way that physical objects appear to observers under standard circum-
stances. For perceptual objects are characterized in terms of sensible
qualities that do not belong to the objective nature of things:
Colors are not properties of the bodies to whose intuition they belong, rather
mere modi¬cations of our sense of sight, which is a¬ected in a certain way by
light. Taste and color are not necessary conditions under which objects can
alone become objects of the senses. They are only connected to appearance as
contingently added e¬ects of [our] particular organization. ( -; see also 

Now association requires not only repetition of patterns, but also
signi¬cant regularities. These regularities are provided by perceptual
objects. But there is some question as to whether the kinds of signi¬cant
regularities of co-occurrence provided by perceptual objects must not
themselves be underwritten by the kinds of laws governing physical
objects, that is, governing empirical objects as they are in themselves. To
say that an object is thus and such is to commit ourselves to the fact that
the object must be representable in that way from all standpoints, even if
it is a contingent fact that it has those standpoint-independent proper-
ties. The object is what Kant refers to in the Transcendental Aesthetic as
the physical or empirical (as opposed to the transcendental) notion of a
thing in itself: ˜˜what in universal experience amongst all the di¬erent
positions relative to the senses, is still thus and not otherwise deter-
mined™™ ( / ). The empirical notion of an object is just the notion
of something in our experience that is what it is independently of my
subjective manner of representing it; the object is thus determined for
absolutely all of us in one way rather than another way. The laws
governing the way such an object presents itself to any arbitrary ob-
server would support the counterfactuals needed to spell out the circum-
stances under which something appears thus and such to standard and
non-standard observers, and hence explain why certain patterns recur
and others do not.

 ®°  µ ¬    §® © ©® ®¤ ¬ ·
Kant introduces recognition as an adjunct to reproduction. Recognition
is the consciousness that the data retained from past experience in one™s
Concepts, laws, and the recognition of objects
present consciousness is in fact the same data as that previously experi-
enced. Without recognition, there would be no experience or memory
in the sense with which we are generally familiar. Memory involves not
only the ability to call up past data, but also the ability to recognize such
data as derived from past experience. Kant notes that retention of past
facts and the recognition of those past as what they are, are distinguish-
able capacities.
The recognition of something as something, for instance, of a past
fact as a past fact, has a normative dimension to it. Recognition can be
successful or unsuccessful. Kant argues that recognition depends on our
ability to use concepts. These concepts supply the norms governing the
distinction between successful and unsuccessful recognition of a certain
object. Concepts serve as rules for organizing the data of intuition by
providing us with the capacity to recognize items in di¬erent situations
( ±°). Kant does not just think of a concept as ˜˜according to its form
always something universal, and that serves as a rule,™™ he also argues
that ˜˜it can only be a rule of intuitions by: representing the necessary
reproduction of the manifold with respect to given appearances, that is,
the synthetic unity in their consciousness™™ (A ±°). This means that
someone who recognizes something as an instance of a certain concept
will necessarily connect the possession of certain properties by that
object with the possession of other properties. For, to fall under a
concept, an object must have certain features for all persons who are
competent with respect to that concept. The necessity relation extends
not only to the relations between the object and the properties that are
represented by a given concept, but also to the relation between the
concept and the object itself. If the object does not fall under the concept
with necessity, but merely accidentally, then, on Kant™s view, the con-
cept is not one that governs our understanding of that object.
By ascribing universality and necessity to concepts in the way that he
does, Kant thinks of concepts as laws. For he thinks of concepts as rules
that associate contents of representation in a necessary way, and he
notes that ˜˜rules insofar as they are objective (hence necessarily attach
to the cognition of an object) are called laws™™ ( ±). There are two
obvious objections to the view of conceptual recognition that Kant
defends in this context. One obvious problem is that the view seems to
entail that all concepts are a priori, for Kant maintains that ˜˜necessity
and strict generality are certain indications of cognition a priori, and
also belong indivisibly together™™ ( ). The other related problem is that
taking necessity and strict generality to be marks of a concept does not
µ Kant and the demands of self-consciousness
seem to do justice to the di¬culties in de¬ning concepts to which Kant is
otherwise quite sensitive: ˜˜One uses certain marks only so long as they
are su¬cient to distinguish; new remarks take some away and add some,
the concept therefore never stands within certain limits™™ ( ·/  ·µ).
The concepts that we have purport to represent kinds, but they may
not. Concept formation is however guided by the regulative ideal of
representing the world in terms of natural and arti¬cial kinds that must
have the same properties for all observers who succeed in identifying
them. Kant wants to argue that our knowledge of such kinds is only
possible to the extent that we are committed to the existence of a priori
laws that make it possible for us to seek and identify kinds of objects.
While he wishes to argue that the lawlikeness of the connection between
the properties belonging to objects of individual kinds is not intelligible
independently of a priori laws, he also wishes to claim that what natural
and even arti¬cial kinds there are is a matter of empirical discovery.
Since laws are not mere regularities for Kant, but necessary and strictly
general or universal principles, they cannot be known by induction on
past experiences alone. They thus presuppose some a priori principle
that allows us to know that they apply in every situation and with
necessity. This view is not now popular; however, it does have in its
favor that there are signi¬cant di¬culties with inductive generalizations,
especially if we are unable to appeal to a principle of the uniformity of
nature that cannot be justi¬ed from past experience, for what is at issue
is whether future experience must be like past experience.

  ®   ° µ  ¬     § ® ©  ©  ®  ® ¤   µ  
Kant links his account of conceptual recognition to an account of the
conditions under which cognition of an object is possible. He interprets
the notion of a concept in terms not only of a necessary rule for
recognizing items in di¬erent experiences, but also of the notion of a
cognition or knowledge of an object (Erkenntnis eines Gegenstandes). Ac-
cording to his more inclusive de¬nition of cognition or knowledge
(Erkenntnis) in the ladder of di¬erent representations in the Transcenden-
tal Dialectic, any representation of an object is a cognition. Thus, not
only judgments, but even intuitions and concepts are instances of
cognition, since they are ˜˜objective perceptions™™ ( °/ ··), that is,
representations of objects. Even this expansive usage leaves a distinction
to be drawn between a representation that is and a representation that
only seems to be true of an object. For we do not want to say that
Concepts, laws, and the recognition of objects
representations must be true of their purported objects. Thus, we still
have room for the idea that in recognition we can either get our
recognition right or not.
Unfortunately, we have no direct way of comparing our concept with
its purported object. We could use another representation, but then the
question would arise again with respect to the new representation. We
need some non-contingent or necessary relation of correspondence
between a representation and its object to support cognition because the
only access we have to the necessary connection between a putative
cognition and its object is through our beliefs about how objects are
connected. Now we can appeal to the way the object must appear to
everyone as a basis for claiming that the object is truly thus and such.
But we can only determine how the object must appear to everyone to
the extent that we have some way of comparing and contrasting the
beliefs that we have about the object with the beliefs that other persons
in other situations have about it. Kant argues that our only ultimate grip
on the purported correspondence of our cognitions with their objects is
based on the coherence relations between those cognitions themselves.
Moreover, these coherence relations consist in nothing but our ability to
connect those cognitions together in ˜˜formal unity of consciousness™™
provided by our capacity to represent the numerical identity of the self
through the changes in representational content involved in represen-
ting these di¬erent cognitions.
To distinguish an object distinct from our particular perceptions, we
need the kind of ˜˜objective unity™™ provided by concepts of space and
time. These concepts allow us to connect inherently perspectival per-
ceptions together in a representation of space and time that is shared,
public, and standpoint-independent. But this shared and standpoint-
independent representation of space and time is nothing but a unity of
consciousness that is necessary in the sense that it must potentially
include all of our possible representations in it. It is on the basis of this
potential unity of consciousness that empirical concepts then have
objective reality, their purchase on objects of experience ( ±°).
In appealing to an underlying necessary connection between the
items that one experiences, Kant is responding to a skeptical dilemma
concerning truth that he posed in the Introduction to the Transcen-
dental Logic ( µ·¬./ ¬.). After noting that the skeptic can accept
the nominal de¬nition of truth as correspondence of knowledge to its
object, Kant points out that this does nothing to resolve the famous
skeptical problem of the criterion of truth. It does not seem to be
µ Kant and the demands of self-consciousness
possible to provide a criterion of truth in a completely general form,
since one would then have to abstract from all content. A logical
criterion of truth, at best, is a necessary condition for non-logical truth.
Thinking that logic alone is su¬cient to establish substantive truths
leads to (dialectical) illusions of the kind that make a critique of pure
reason necessary. Despite the tendency of reason to make substantive
claims that it cannot substantiate, Kant thinks that this tendency can
be curbed by limiting cognitive claims to experience and its precondi-
tions. The Transcendental Analytic (i.e. the Transcendental Deduction
and the Principle of Pure Understanding) constitutes a ˜˜logic of truth™™
in contrast to the ˜˜logic of semblance™™ provided by the Transcendental
Dialectic. The Analytic provides a set of substantive necessary criteria
(where a criterion is a canon or measuring stick) for empirical truth, but
not for objects in general ( “/  ·“). Our beliefs can be
determined to be true or false depending on whether they can be
connected together in one consistent account of experience. The con-
nectability of our beliefs gives us a coherence criterion of truth. As
Kant notes in the Appendix to the Transcendental Dialectic, without a
˜˜use of the understanding™™ that ˜˜hangs together™™ (zusammenhangenden ¨
Verstandesgebrauch) ˜˜we would have no su¬cient mark of empirical
truth™™ ( µ±/ ·). But we cannot determine the truth of claims that
go beyond what we can experience.
While the combination of a coherence criterion of truth with a
correspondence de¬nition of truth has much to be said for it, it is far less
obvious why the connections between our perceptions, or beliefs in
general involved in the establishment of coherence, must be necessary
and strictly general. The reason to make such a demand lies in the idea
of a general criterion of truth. If a criterion of truth is to tell us what
beliefs are true or false, then it must pick out all beliefs that are true. This
leads Kant to conclude that our beliefs must not only have necessary
connections amongst themselves if they are to have the necessary
connection to an object that he believes characteristic of knowledge,
these necessary connections must also be based on a priori rules.
In arguing that conceptual recognition is required if we are to be able
to make sense of experience, Kant is most interested in the status of
non-empirical concepts. For it is the task of the Deduction to show that a
certain set of such concepts, the categories, or pure concepts of the
understanding, are legitimate concepts that pick out bona ¬de objects.
To do this, Kant wants to show that these a priori concepts are
necessary if we are to be able to recognize objects:
Concepts, laws, and the recognition of objects
Real experience, which consists in apprehension, in association (reproduction),
¬nally in recognition of appearances, contains in the last and highest (of the
merely empirical elements of experience) concepts that make possible the
formal unity of experience, and with it, the objective validity (truth) of empirical
cognition. Now these grounds for recognition of the manifold, insofar as they
concern merely the form of experience in general, are those categories. ( ±µ)

It is clear from the above passage that Kant thinks of the categories as
the conditions that make it possible for us to distinguish a successful
from an unsuccessful recognition of an object, that is, to distinguish a
true from a false judgment about an object. How do they allow us to
draw the distinction in question? They allow us to distinguish a unity of
experience for me, from a unity of experience for everyone by establish-
ing constraints on what counts as the form of experience in general. The
key thing for Kant is that in order for there to be a successful conceptual
cognition there must be necessary reproduction, that is, association
according to an a priori rule ( ±°µ). The a priori rule in question cannot
itself be the concept through which we recognize what we have asso-
ciated together in the mind, for then all concepts and all cognition
would be a priori. The idea is rather that, in order for one to be able to
apply even empirical concepts to experience, there must be necessary
and strictly general connections between items of experience. It is this
metarule that must be a priori. The metarule ¬xes the relations of
potential data for knowledge a priori in such a way that empirical or
non-empirical rules can be found that are capable of supplying us with
Ultimately, Kant wants to argue that even the metarules, or catego-
ries of the pure understanding, that provide necessary and strictly
general connections between items of experience derive their normativ-
ity from the unity of apperception itself:

The unity of apperception is the transcendental ground for the necessary
lawlikeness of all appearances in one experience. And this very unity of
apperception in respect to a manifold of representations is the rule (namely to
determine it [the manifold] out of one [representation]) and the faculty of these
rules is the understanding. ( ±·)

 ° °    ° © ®  ®¤   ®    ® ¤  ®  ¬  ¦ ¦ © ® ©  
In arguing that association requires laws governing physical objects that
are independent of our immediate experience of those objects, Kant is
µ Kant and the demands of self-consciousness
interested in the nature of the similarities and uniformity required for
associative processes of the kind postulated by Hume to get o¬ the
ground. He wants to show that lawlike regularities underwrite the
mechanisms of association that Hume regarded as ˜˜the cement of the
Hume admits that appearances cannot suggest rules for projecting
predicates, unless there are constant conjunctions of appearances in our
experience. Without such constant conjunctions, associative processes
could not be initiated. For Hume, such patterns of association are based
on innate dispositions to associate recurrent sets of appearances with
one another. But even these innate dispositions must be triggered by
similarities in what is perceived. Hume™s e¬ort to base causal connection
on constant conjunctions is thus sensitive to the conditions under which
events are to be regarded as similar. Consonant with his empiricism,
Hume attempts to derive the principle that every event has a cause from
the principle of like cause“like e¬ect. Since he eschews a priori knowl-
edge, Hume must treat the universal causal principle as a consequence
of our custom of treating nature as uniform. The uniformity of nature in
turn falls out of our habit of associating similar causes with similar
Kant wants to argue for the thesis that ˜˜the order and regularity in
the appearances, which we entitle nature, we ourselves introduce. We
could never ¬nd them in appearances, had not we ourselves, or the
nature of our mind, originally set them there™™ ( ±µ). Although the idea
that we somehow produce the order of nature is controversial to say the
least, there is nothing in this general idea that a Humean would have to
reject. Kant does, however, soon give the thesis that we impose order on
nature an interpretation which Hume could not accept: ˜˜For this unity
of nature has to be a necessary one, that is, has to be an a priori certain
unity of the connection of appearances™™ ( ±µ).
We must think of nature as a whole as conforming to the conditions
under which we can make nature intelligible to ourselves. This is what
gives nature its unity for us. But then, the only access to nature we have
is through what we represent of it. Since nature is accessible to us only
insofar as it conforms to the conditions under which we can make sense
of nature, the nature which is accessible to us is nature as it must appear
to us. Nature which appears to us is subjective only in the sense that it is
construed as dependent on the conditions under which we can make
sense of it. These same conditions, however, also serve to underwrite the
objectivity and accessibility of our claims concerning nature for other
Concepts, laws, and the recognition of objects
persons. Kant refers to the unity and uniformity of nature as nature™s
a¬nity. He distinguishes between an empirical and a transcendental
a¬nity of nature. In other words, there is thus supposed to be an
empirical and a non-empirical dimension to the uniformity of nature.
The empirical a¬nity of nature consists in the uniformity that we
discover in nature based on the kinds of associations that we make
between di¬erent items of experience based on repeated experience.
Transcendental a¬nity, by contrast, is supposed to be the underlying
uniformity of nature that makes it possible for us to associate items
together in a meaningful manner in the ¬rst place. Kant™s claim is that
uniformity of nature, in the transcendental sense, is based on our
capacity to connect di¬erent representations together in (transcenden-
tal) self-consciousness.
Kant starts with the empiricist notion that we become conscious of,
and indeed construct, bodies and selves through the manner in which
we associate representations, or rather simple sensations. Then he notes
that association of representations means forming some kind of connec-
tion between those representations in consciousness. He then argues
that any connection that might hold between individual representations
in a particular consciousness must be such that it is consistent with the
uni¬ability of those representations in self-consciousness. The move
from an empirical a¬nity or associability of representations in an
individual empirical consciousness to the idea that empirical associ-
ations are governed by a priori principles of associability is based on the
idea that any perceptions that can be connected in any particular
consciousness must conform to the conditions under which they can be
associated with other representations in an arbitrary consciousness. The
idea of the universal associability of representations in one possible
self-consciousness is then all that an object that exists independently of
my particular associations could be.
Henry Allison has criticized Kant™s argument for taking a tack similar
to Berkeley™s notorious argument to the incoherence (˜˜repugnance™™) of
the notion of unperceived matter.µ On this reconstruction, Kant argues
as follows: things that appear to me, such as cinnabar, tables, chairs,
etc., exhibit regularities that support associations connecting them to-
gether in various ways. These regularities are necessary. For only if
appearances can be connected together in consciousness, and hence can
be associated, do they exist at all. Something represented by me that
cannot be a possible object of consciousness would be nothing at all.
Therefore all appearances must be necessarily associable. The argu-
µ Kant and the demands of self-consciousness
ment involves a fallacious shift between two di¬erent meanings of
˜˜appearance,™™ corresponding to di¬erent senses in which consciousness
can be unitary. An appearance in the subjective sense, an appearance±,
is an internal accusative of consciousness (a represented qua represen-
ted). It would be nothing at all, if it could not be represented. An
appearance in the objective sense, an appearance, or a phenomenon in
Kant™s technical vocabulary, is an intersubjectively accessible object.
This is an object that is independent of any particular mind or spatio-
temporal perspective. If one thinks that appearances± are necessarily
associable, then we may infer the existence of appearances that under-
lie appearances.±. But what reason does one have to think that appear-
ances± must be representable together by a subject?
Representations are entities that play a role in understanding the
behavior of animals and human beings. If such entities are assigned a
role that makes them cognitively meaningless, then there is no reason to
assume their existence at all. The key claim however is that, in order for
such representations to be cognitively signi¬cant, they must have di¬er-
ential relations to other representations that we can represent to our-
selves. Appearances must provide some ground for the association of
appearances±. This is what it is for them to be in themselves associable
and thus susceptible to conceptual interpretation. Without this objective
basis for associability, one would not be able to conceptualize the
episodes in question and thus one would not be able to represent oneself
as a numerically identical subject through these di¬erent experiences. It
is only if those experiences allow for the application of concepts to them
that it is possible to represent oneself as the numerically same subject
with respect to them.
Similarity relations between concepts are themselves to be under-
stood in terms of their systematic relation to other concepts and repre-
sentations in a possible self-consciousness. Not all di¬erences in repre-
sentational content can come from experience itself, since there must be
some way of thinking of the relations between representations that is
independent of any particular experience, if we are to be able to
represent their similarity relations as such at all. We must think of our
experience, both representations and their objects (what is represented
by them), as conforming on the whole to the conditions under which we
can apply such concepts to experience, since this is the only way we have
of making sense of experience, and giving any determinate content to
our representations or thinking of ourselves as using concepts and
forming judgments:
Concepts, laws, and the recognition of objects
All error into which the human understanding can fall is only partial and in
every erroneous judgment there must be something true. For a total error would
be a complete contradiction against the laws of the understanding and reason.
How could it arise as such in any way from the understanding, and insofar as it
is a judgment, be held for a product of the understanding! (Logic, ed. Jasche, Ak.
©, p. µ)
We can be mistaken in particular beliefs or representations of objects,
but the very conditions under which we can make sense of the content of
our representations precludes us from being completely wrong in all of
our beliefs. The content of our representations (our beliefs) is linked to
what is intelligible for us. In this way global skepticism is rejected.

 µ®©¦  © ® ¤       © µ ®© ¦ ®  µ 
Kant™s argument to the transcendental a¬nity of nature suggests that
nature must be uniform in every respect if there is to be self-conscious-
ness. He wants to claim that there is su¬cient uniformity in cinnabar
with respect to its properties for us to have the notion of how cinnabar
looks to standard observers under standard circumstances. Deviation
from standard conditions will be made sense of in terms of a change in
causal conditions. He assumes the validity of the prediction that, if
cinnabar was red in the past, then it will be red in the future, since the
future resembles the past. But, of course, this can mean at best that
cinnabar has the dispositional property of appearing as red to standard
observers under standard conditions. Such conditions will be deter-
mined in causal terms. Cinnabar is a di¬cult case because it may be
observed to be both red and black depending on other causal factors
such as oxidation. Its redness and blackness are dispositional properties
connected both to the absence or presence of oxygen and to the manner
in which our eyes process radiation of a certain wavelength.
One could take nature to be uniform with respect to some Goodman-
type predicate such as the ˜˜brackness™™ of cinnabar. If something has the
property of brackness then it will be red before ±°° and black after that
date. It would be contradictory to say that nature is uniform with respect
to both of these properties. Nature cannot be uniform in respect to every
concept either. If nature is uniform with respect to the concept ˜˜red,™™
we will not be able to think of it as uniform with respect to the concept
˜˜gred™™ and vice versa, where the latter concept is to be understood as
˜˜being red prior to ±·· and being green thereafter.™™ For, if we take the
predicate ˜˜gred™™ as basic, we may then de¬ne ˜˜red™™ as ˜˜gred before
° Kant and the demands of self-consciousness
±·· and redd afterwards,™™ where redd would be a predicate de¬ned in
the red vocabulary as ˜˜red after ±··.™™ Since mutually inconsistent
universal conditional hypotheses involving both normal predicates and
Goodman predicates can be formulated which are equally supported by
available evidence, nature cannot be taken to be uniform in every
respect. Some constraints are needed which will allow one to distinguish
legitimate from pseudo-lawlike judgments. Even if experience exhibits
regularity with respect to some concept or other, this is not enough to
guarantee regularity under laws. More needs to be said about the kinds
of regularity to be allowed.
Predicates are projectible, if they allow projection from observed to
unobserved cases. Redness rather than brackness belongs to the group
of predicates that we hold to be projectible on intuitive grounds. Red-
ness is more deeply entrenched in our belief system. Projectible predi-
cates are predicates that ¬gure in generalizations expressing lawlike
connections between events rather than merely accidental regularities
of co-occurrence. Such predicates allow the universal conditional hy-
potheses containing them to be con¬rmed by positive instances. This
means that the truth of singular empirical judgments that are instanti-
ations of those hypotheses increase their credibility. Projectible will need
to be distinguishable from non-projectible predicates. The uniformity of
nature will also have to be restricted to projectible predicates. Otherwise
there will be no fact of the matter about whether there is a change in
state of some substance from time t± to t.
If cinnabar is brack, then the change from redness to blackness at
t=±°° will not be a true change, since it is not a change in the brackness
of that substance, whereas the persistence of the color red after that date
would actually require further explanation, since it would involve a loss
of brackness. It is certainly tempting to exclude Goodman predicates
based on the Kantian demand for temporal neutrality. But, as the
reference to causation and change suggests, the demand works both
ways. From the point of view of the vocabulary using brackness, redness
would itself need to be de¬ned in temporal terms, i.e. as the having of
the property of brackness prior to ±°° and of another property which
may be called wreckness after that date.
In order to respond to the challenge of Goodman predicates, Kant
must establish some basis for a priori constraints on which concepts can
serve as concepts of a possible experience. Only in this way will he be
able to isolate the causal conditions governing particular experiences in
any plausible way. Kant, in fact, has the sketch of an a priori answer to
Concepts, laws, and the recognition of objects
this kind of problem. He demands systematic uni¬cation of concepts as
a regulative principle governing universal conditional hypotheses of the
kind which play a role in theories. Systematic uni¬cation should exclude
Goodman predicates by restricting the range of acceptable predicates to
projectible predicates. Our choice of projectible predicates must be
justi¬ed by reference to the best background theory that we have
available to us. Our background theory will have to be not only strongly
supported by available evidence, but also constrained by regulative
principles governing the best way of systematically unifying the informa-
tion gleaned from this evidence in a theory.
Non-natural Goodman-type predicates cannot be ruled out on an
individual basis. But, if we think of our concepts as possessing constitut-
ive systematic relations to each other, then it is plausible to assume that
they must confront experience as a systematic totality, rather than on a
one-by-one basis as Humeans are inclined to believe. Hume™s account
of induction is not, in fact, up to drawing plausible distinctions between
natural and non-natural predicates. The problem is that Hume would
have such predicates depend for their support on present experience of
constant conjunctions of perceptions. But there will always be non-
natural predicates that can ¬nd experiential support equal to that to be
had for so-called natural predicates. These non-natural predicates can
only be excluded by rejecting a molecularist account of concept applica-
tion. The problem with non-natural predicates is that they cannot be
systematically deployed and still maintain their non-natural meaning.
One can ¬nd non-natural predicates corresponding to any given set of
natural predicates, but one cannot construe all concepts as having such
non-natural meaning.
Concepts are systematically related to each other in a manner that is
both parasitic on the numerical identity of the self and a condition for its
possibility. Thus, if self-consciousness is possible in experience, then one
must conceive of experience as subject to concepts whose content is
systematically interrelatable and relatable to that self-consciousness.
This connection between the systematic unity of concepts and the
identity of self-consciousness lies behind Kant™s interpretation of the
di¬erences between concepts on the model of di¬erences in the content
of what is represented by a spectator from di¬erent standpoints:

One can regard every concept as a point, which, as the standpoint of a
spectator, has its horizon, that is, a set of things, that can be represented and as
it were surveyed from it. But to the di¬erent horizons, that is, kinds, that are
 Kant and the demands of self-consciousness
determined by that same number of concepts, one can think of a shared
horizon from which one observes them all from a point in the middle which is
the higher kind, until ¬nally the highest kind is the universal and true horizon
which is determined from the standpoint of the highest concept, and contains
all the manifold under it as kinds, species, and subspecies. ( µ/ )
Our representations and, more directly, our concepts ¬gure in a
systematic unity that relates our concepts to each other systematically.
Systematic uni¬ability is an a priori constraint on experience. From
Kant™s point of view, inquiry is guided by the assumption that nature is
characterized by a transcendental a¬nity among its concepts. Nature is
transcendentally a¬nite with respect to concepts insofar as nature
allows itself to be understood in terms of a system of concepts. We must
think of nature as conceptualizable in terms of a system of concepts
which could, in principle, be articulated by us in some systematic way to
the extent that we think of nature as something that we can represent in
terms of a self-consciousness that is basically impersonal.
While the transcendental a¬nity of nature is, in many ways, the
linch-pin of Kant™s argument for the a priori validity of the categories, it
seems to be describable with equal justi¬cation as a transcendental
deduction of the synthetic a priori maxims of unity, multiplicity, and
relatedness or a¬nity. These ideas of reason provide the transcendental
justi¬cation for the assumption that the complex conceptual hierarchies
to be found in highly articulated theories are indeed applicable to
nature. The hierarchical organization of concepts in a theory allows
every event to be brought under not only a true generalization, but also
a generalization that will be helpful in making predictions about other
events. In fact, the problem of uniformity under concepts and laws
reveals the close connection that must be demanded between under-
standing and reason.
The A-Deduction requires the uniformity of nature and, indeed, the
lawlikeness of nature for any understanding or recognition of the world
as what it is. But by the time one reaches the Transcendental Dialectic (
/ ·), it becomes apparent that any attempt to formulate the unity
and laws of nature is necessarily hypothetical. The attempt to formulate
an abstract conception of nature that is uniform with respect to all
conceivable concepts generates antinomies. We do have to think of
nature as conceptualizable by us, and hence as organized into a system
of natural kinds that we can understand. But we cannot know a priori
what that system must be like in its particulars, since there are a number
of equally plausible, but mutually exclusive, alternatives. This does not
Concepts, laws, and the recognition of objects
mean that there must be mutually exclusive empirically adequate sys-
tematic characterizations of nature. For we do not have su¬cient
evidence to do more than provide a projection of what an empiricaly
adequate system of the world would be like.
In the Appendix to the Transcendental Dialectic, Kant regards the
notion of a¬nity primarily as a property of concepts rather than as a
property of their objects (that is, of the manifold itself ), as he does in the
Deduction. But the notion of a¬nity developed in the Appendix has
transcendental rather than purely logical force. It expresses a transcen-
dental assumption to the e¬ect that nature is susceptible to explanation
in terms of the hypothesis of conceptual relatedness. Thus, the a¬nity in
question is ¬rst of all constituted by a set of similarity relations between
concepts, but these similarity relations between concepts must also be
relevant to the objects that are interpreted in terms of those concepts,
and hence to what Kant calls the manifold. This is brought home
forcefully in the following passage:
If there were such a great di¬erence among appearances, I do not want to say of
form (for in this respect they may be similar to each other), but with respect to
content, i.e. with respect to the multiplicity :Mannigfaltigkeit9 of existing
beings, that the sharpest human understanding could not ¬nd the least similar-
ity between them by comparing the one with the other (a case which is
conceivable), then the logical law of kinds would not take place at all, nor even a
concept of a kind, or any general concept of any kind, indeed there would not
even be understanding which only has to do with such. The logical principle of
kinds presupposes therefore a transcendental one, if it should be applied to
nature (by which I understand only objects that are given to us). ( µ“µ/
The transcendental principle of similarity between kinds seems to be
precisely what Kant needs in order to be able to claim that association is
based on some principle which makes occurrences associable in them-
selves. This transcendental principle of similarity seems to have its
ultimate source in the fact that we must regard all representations as
potential candidates for self-consciousness.
The systematic interest of reason in its pursuit of uni¬cation is limited
in its empirical signi¬cance to the conditions governing the legitimate
acquisition and application of empirical concepts. But even Kant thinks
that the attempt to come up with precise a priori constraints on which
concepts can function as concepts of empirical objects overtaxes both
the understanding and reason. Each attempt to provide a speci¬c
understanding of the a¬nity of nature is necessarily hypothetical. It
 Kant and the demands of self-consciousness
must be based on a projected unity of nature that transcends what we
know through experience even though it is subject to revision in the face
of recalcitrant experience.· The hypothetical character of any unity of
nature demotes the synthetic principles a priori that function as criteria
for the truth of our empirical judgments to the status of regulative rather
than constitutive principles of experience. This is true even though we
are unable to apply concepts without the background assumption about
experience with which they provide us. Kant insists, to be sure, that the
law of reason, that one must look for unity in nature according to
principles of reason, is itself necessary:
Without it [the law of reason] we would have no reason, without reason no
coherent use of the understanding, and without that no su¬cient condition of
empirical truth, so that in respect to the latter we must presuppose the
systematic unity of nature indeed as objectively valid and necessary. ( µ±/
A proof of the transcendental a¬nity of nature is involved in Kant™s
justi¬cation of the validity of the categories. For these categories are
supposed to provide a general notion of the lawlikeness of nature and its
susceptibility to interpretation by means of empirical concepts. But,
even were this project completely successful, there would still be room
for the use of the organizational powers of reason and its ideas in
attempting to formulate more speci¬c models of objective a¬nity and
speci¬c laws of nature. It is these models that then guide the inductive
enterprise of science. The principles that reason formulates are indeed
synthetic and a priori, but they do not determine the character which
objects of experience must have. They cannot tell us how individual
objects of experience must be, but principles of reason do have an
objective validity that transcends mere thought-economy. They allow us
(provisionally) to determine whether our representations are true of the
objects of which they seem to be true. In regulating our search for
empirically adequate lawlike connections between natural objects as
well as between natural objects and their properties ( / ±),
principles of reason provide substantive constraints on which of our
representations can be true. Principles of reason are thus crucial in
providing the necessary connectedness among our experiences that
Kant takes to guarantee the correspondence of our representations with
their object, that is with an empirical reality that is distinct from those
representations ( ±°µ).
° 

Self-consciousness and the demands of judgment
in the B-Deduction

Kant™s most sophisticated treatment of how self-consciousness con-
strains the character of experience is to be found in the B-Deduction. In
the A-Deduction, in which the notion of judgment is only mentioned
once in describing the powers of the understanding ( ±), Kant™s
argument turned on the enabling conditions of recognition. The B-
Deduction establishes objectivity by way of a more explicit appeal to the
normative demands placed on experience by the possibility of forming
judgments about what is experienced. Kant ¬rst argues that any cogni-
tively signi¬cant content is a potential candidate for representation in a
consciousness of self that potentially includes all representations whatso-
ever. He then argues that whatever is a candidate for self-consciousness
is also something to which we can apply concepts and hence a candidate
for judgment.
Judgments make an implicit claim to objectivity by making a truth
claim. In forming a judgment, we commit ourselves to the truth of the
proposition that is asserted by the judgment. It might be thought that
this truth could merely be a truth for me or for someone else. In this
case, the truth would be merely subjective. However, such a subject-
relative conception of truth would not do the job that we assign to the
notion of truth, namely to capture the way things are independently of
an individual point of view or take on the way the world is. For this
reason, Kant accepts the nominal de¬nition of truth as correspondence
with an object even though there is no way to determine whether a
judgment corresponds to an object independently of whether that
judgment coheres with other judgments.
The claim to truth made by judgment, and with it the presumption
that the proposition asserted by the judgment corresponds with an
object, is the ground for the normative claim made by a judgment. This
normative ground of judgment ultimately has its source in the possibility
of representing the content asserted by a judgment in an impersonal
 Kant and the demands of self-consciousness
way. The possibility of representing the contents of judgment imperso-
nally, in turn, is based on the fact that consciousness of one™s particular
point of view as a representer is parasitic on the possibility of represen-
ting oneself in an impersonal manner.

˜ ˜ ©  © ® «™ ™ © ®    - ¤  ¤µ   ©  ®
Kant introduces the idea that the unity of self-consciousness is the
normative source of all content that can have any cognitive signi¬cance
for us in a passage that is as famous and controversial as any in his whole
The I think must be able to accompany all my representations; for otherwise
something would be represented in me which could not be thought at all, which
would mean as much as that this representation would either be impossible or
at least nothing for me. That representation which can be given prior to all
thought is intuition. Therefore all the manifold of intuition has a necessary
relation to the: I think in the same subject in which this manifold is to be
encountered. ( ±)
Kant™s initial aim in this passage is to show that I can think all my
representations. He apparently means that all my representations, taken
collectively, as well as distributively, are ascribable by me to me. To
establish the initial conclusion, Kant assumes for the sake of argument
that it is not the case that all my representations are thinkable by me.
From this, he claims that it would follow that something would be
represented in me that could not be thought at all. In other words, he
assumes that, if something in me is to be represented by anyone, it must
be represented by me. From this, he concludes that, if I cannot think a
representation in me, then no one else can either. Although he initially
seems to reject the possibility of representations in me that cannot be
thought, he then goes on to concede that there might be such represen-
tations, although they would be nothing for me.
The existence of thoughts in me that are not thinkable would certainly
entail a contradiction. A thought that could not be thought would be an
impossible representation. But it is not so obvious that I must be able to
think of all thoughts in me as my thoughts. Now Kant does not initially
argue for the claim that all thoughts must be potential candidates for
self-consciousness. One might argue that to have a thought of something
is to judge or at least entertain the possibility that something is the case. It
seems plausible to maintain that one cannot have thoughts in this sense
without being somehow aware that one is having them. However, Kant
The demands of judgment in the B-Deduction
has a less direct and ultimately more persuasive argument linking all
(discursive) thought to self-consciousness. The argument for that claim
comes later. For the time being, I will simply assume that an argument for
the claim that all thoughts are potential I thoughts is forthcoming.
Now one can see that Kant™s principle that the ˜˜I think™™ must be able
to accompany all my representations would be established if there could
not be representations in me that are nothing for me. It is reasonable to
assume that, if a representation in me is to be something for me, it must
be something that I can think of as mine. If only such representations
were possible, it would seem that all representations in me would have to
be thinkable by me. But Kant™s suggestion that there might be represen-
tations in me that are nothing for me seems to throw a monkey-wrench
in the argument. In restricting representations that are mine to those
that are thinkable by me, one leaves open the possibility of representa-
tions other than thoughts in me, i.e. intuitions, that are not thinkable at
all. The point to note here is that the existence of representations that
are nothing for me, representations of which I am not conscious, is
consistent with the principle that I can think all my representations so
long as one restricts the meaning of mine to those representations in me
that are something for me. The validity of Kant™s argument thus
depends on the assumption that representations in me that would be
nothing for me cannot be mine at all.
Robert Howell worries with Paul Guyer that Kant falsely infers that I
can become conscious of each representation that happens to be mine
from the fact that I can become conscious of a representation as mine, if
I represent it as mine.± But the point to note here is that Kant under-
stands mineness in a restrictive sense. What is mine is something for me,
as opposed to something of which I might be an owner in a sense that is
cognitively inaccessible to me. I may own something, even though I do
not know that I own it. I might even own something that contingent
circumstances prevent me from ever recognizing as my own. But if I do
not know that I own something, there must at least be some evidential
base for a possible claim to ownership by me. There might be a
misplaced deed or a long-forgotten relative to support my entitlement. I
could appeal to these sources of evidence for my ownership if I were
aware of them. The same general principle of entitlement that applies to
property applies to the ownership of representations or mental states. If
a representation is mine in principle, I must be able to recognize it as
such, even if a representation can be mine without my being conscious
at that time that it is mine.
 Kant and the demands of self-consciousness
I have interpreted Kant™s argument as turning on a distinction
between a weak notion of existence in my consciousness and a stronger
notion of being mine. By contrast, Henry Allison suggests that there can
be representations that are mine yet are nothing for me. He then argues
that one can take the ˜˜I think™™ principle to be restricted to a subset of
my representations, namely, those that function as representations
through which I represent something to myself. While Allison insists
that representations could be mine in some sense without being some-
thing for me, Kant™s argument depends on the assumption that, if a
representation is mine, it is something for me. Kant allows for the logical
possibility of representations in me that are nothing for me. He claims
implicitly that such representations would not a¬ect the validity of the
claim that the ˜˜I think™™ can accompany all my representations. But he
cannot allow for representations that are mine, but nothing for me.
Kant clearly does assume that representations which happen to be in
me, but that I cannot think of as mine, are without cognitive signi¬-
cance. For, without such an assumption, his argument, establishing the
claim that even representations in me that can be given prior to all
thought must be potential contents of I thoughts, would be invalid. For,
after implicitly distinguishing possible representations in me that are
nothing for me from my representations, and arguing that all my
representations are thinkable by me, Kant notes that intuitions are
representations that are independent of thought. Surprisingly, he goes
on to conclude that all contents of intuition in me must be thinkable by
me, where one would have expected him to conclude that only all the
contents of my intuition must be thinkable by me. The initial phase of
Kant™s argument seems to be free of the con¬‚ation of two di¬erent
senses of ownership: ownership in the sense of what can be ascribed by
me to me, and ownership in the weaker sense of what merely exists in
me. But it is at this juncture in the argument that Henrich™s worry, that
Kant falls victim to an ambiguity in the notion of mineness in extending
mineness to all representations that may be represented in me, seems to
have considerable force. However, I think a charitable reading of
Kant™s argument would be that he wishes to understand intuitions in me
as restricted to those that are mine, that is, those that are something for
me. Henrich would also acknowledge that this reading is the one most
consonant with Kant™s aim in this part of the Deduction, although he
worries that Kant may have been taken in by the potential for confusion
involved in the distinction between a representation being in me and
being something for me.
The demands of judgment in the B-Deduction
Now a number of examples have been adduced to show that we are
sometimes incapable of being self-conscious of a certain representation
that occurs in us. This has often been adduced as evidence that Kant
fails to establish that all my representations are self-ascribable. The most
obvious point to make here is that Kant implicitly distinguishes repre-
sentations occurring in us from representations that are ours in the sense
of representations that have some cognitive signi¬cance for us. How-
ever, the examples in question might suggest that representations may
have cognitive signi¬cance and yet not be self-ascribable. It is this
possibility that I wish to address in discussing examples of states of
consciousness that do not seem to be potential states of self-conscious-
Let me start with Guyer™s example in which another person infers
that I have been dreaming from my rapid eye movements. I do not
remember my dream. Therefore the other person™s inferential evidence
is better than my own introspective knowledge. That person can there-
fore ¬nd representations in me that I do not ¬nd. In fact, it is almost
de¬nitive of a dream state that it is a state of consciousness that is not
also a state of self-consciousness. Hector-Neri Castaneda makes the
same point by appeal to the phenomenon of blind-sight.µ In blind-sight
one perceives things without having the capacity directly to be conscious
of them, so that one can only learn by indirect inference from one™s own
behavior what one has perceived. We might also mention examples in
which the corpus callosum of the brain has been bisected. Such bisec-
tion raises particular problems for the unity of consciousness, since each
of the two hemispheres of the brain is capable of functioning indepen-
dently of the other. Terence Wilkerson notes the more familiar
examples of babies and comatose individuals. They have representa-
tions, but they are not self-conscious or even, strictly speaking, capable
of self-consciousness.· From examples such as these, it is reasonable to
infer that there is consciousness without self-consciousness.
Manfred Baum attempts to respond to the possibility of unconscious
representations that might never become conscious to me by arguing
that the only access to my representations is a ¬rst-person access. He
infers from the privileged access that each of us has to our own represen-
tations that there cannot be representations that cannot become con-
scious to the person who has them. Yet, even if Baum is correct that I
have no direct experience of other individuals™ mental states, I may be in
a better position than they to determine what their mental states are. So
privileged access does not rule out the existence of states that can never
·° Kant and the demands of self-consciousness
become conscious to the person that has them. My inferences based on
indirect evidence of their mental states may be less subject to error than
their ascriptions of representations to themselves based on the immedi-
ate data of their experience. Introspection needs correction and ampli¬-
cation from the third-person point of view. On the other hand, we may
grant the assumption that behavior such as rapid eye movements
provides a good inferential basis for the belief that a person is dreaming,
and thus experiencing a certain representation, without granting the
very strong asymmetry between self-ascription and other-ascription to
which Guyer, for one, is committed. If you can have inferential knowl-
edge that I am dreaming through observation of my rapid eye move-
ments, there is no principled reason why I cannot also come to have
inferential knowledge of this as well. I might see myself later on ¬lm and
conclude that I was dreaming.
Kant allows for representations of which we are not directly con-
To have representations and not to be conscious of them, seems to contain a
contradiction; for how can we know that we have them if we are not conscious
of them? Locke already made this objection and therefore also rejected the
existence of this kind of representation. “ But we can become conscious of a
representation indirectly [mittelbar], although we are not immediately aware of it.


. 2
( 9)