A Kantian Critique of Current Assumptions about Self-Knowledge
Patricia Kitcher
Columbia University
1. Three Assumptions about Self-Knowledge
One theme of contemporary discussions of self-knowledge is that where Descartes
and Locke offered seductive and erroneous pictures of the phenomena, Kant’s
approach was minimalist and reasonably accurate. Descartes is an obvious target of P.
F. Strawson’s ‘Persons’ (1959) and of Gareth Evans’s (1982) seminal discussion of selfidentification; Locke’s theory of inner sense has been subjected to withering criticism in
an important series of papers by Sydney Shoemaker.1 By contrast, Strawson, Evans,
Shoemaker and many others align either their negative or positive claims about selfknowledge with Kant’s admittedly enigmatic remarks about the ‘I-think.’ I believe that
the current consensus is only half right. Kant’s exploration of the necessary conditions
for cognition gave him a far better understanding of self-knowledge than his
predecessors. His insights into the topic are not, however, perfectly aligned with current
views. A closer look at the theory of the transcendental unity of apperception suggests
that three foundational assumptions of current work on self-knowledge are importantly
mistaken. Although I have cast the discussion in terms of the history of philosophy, my
arguments will be mainly systematic. I am not going to argue for a reading of Kant, but
from a reading of Kant to a critique of widely held assumptions about self-knowledge.
Still, I think the Kant I present will be sufficiently recognizable to establish my thesis
about the ambivalent relation between his historical doctrines and contemporary
approaches.
The idea of turning to Kant for enlightenment about the self and self-knowledge
comes from the work of P. F. Strawson. Strawson casts ‘Persons’ in terms of finding an
appropriate middle ground between the Cartesian view that
1
when we speak of a person we are really speaking of two distinct substances…each of which
has its own appropriate types of states and properties (1959, 94)
and the no-self view that states of consciousness do not belong to anything. He is
hesitant about attributing the latter view, but he suggests Schlick, Wittgenstein and
Lichtenberg (as well as Hume) as probable proponents (1959, 94-95, 95n.1). He saw
Kant as more perspicuous than Hume because he recognized that
the unity of the ‘I think’ that accompanies all my perceptions and therefore might just as well
accompany none …is a purely formal (‘analytic’) unity. (1959,103)
Strawson’s means for steering between the unattractive alternatives is to work out
the implications of what he took to be two obvious facts:
It is a necessary condition of one’s ascribing states of consciousness, experiences, to oneself,
in the way one does, that one should also ascribe them, or be prepared to ascribe them, to
others who are not oneself.
The method of verification [is] so different in the two cases [the first person case and the
second/third person case]—or rather … there [is] a method of verification in one case (the case
of others) and not, properly speaking in the other case (the case of oneself). (1959, 99, 99-100)
These two claims are then deployed in his admirably succinct account of his central
argument:
One can ascribe states of consciousness to oneself only if one can ascribe them to others. One
can ascribe them to others only if one can identify other subjects of experience. And one
cannot identify others if one can identify them only as subjects of experience, possessors of
states of consciousness. (1959,100).
From this reasoning, it is but a short step to Strawson’s conclusion that ‘person’
is a primitive term, a term which must be used in conjunction with both material and
psychological predicates. Over the years, this argument has prompted many
objections, with the most devastating criticisms focusing on its apparently verificationist
premise that there must be logically adequate behavioral criteria for ascribing
psychological predicates (1959, 106) and hence for knowledge of other minds or
persons.2 Although few, if any, philosophers now accept the argument of ‘Persons,’
many assume that Strawson identified a crucial desideratum for any adequate theory of
2
self-knowledge: It must deal with the very different ways that humans have of acquiring
[not ‘verifying’] self-knowledge and knowledge of other minds.3 I argue below that, from
a Kantian perspective, the difference between knowledge of self and others is fairly
trivial. What is far more important with respect to ‘other minds’ is that subjects can
understand others as subjects only through their understanding of themselves as
subjects.
Gareth Evans follows Strawson’s lead in developing an account of self-reference
that applies equally to bodily and to mental states partly as an antidote to the powerful
Cartesian conception of the self (1982, 220). He also offers subtle and original
accounts of the reference of ‘I’ and of the self-knowledge of beliefs. In his view, the best
linguistic analog for ‘I’ is ‘here.’ He argues that a person succeeds in referring to a
particular location in public space by ‘here’ through a network of dispositions
to direct his actions to that place, and to treat perceptions of that place as germane to the
evaluation and appreciation of the consequences of the thought …about a position in egocentric
space (1982, 168).
He believes that ‘I’ works in a similar way:
It is essential, if a subject is to be thinking about himself self-consciously that he be disposed to
have such thinking controlled by information which may become available to him in each of the
relevant ways. (1982, 216)
For example, someone who uses ‘I’ must be disposed to have his ‘I-thoughts’ controlled
by his perceptual or sensory information, so that seeing red or being in pain can lead
one to say that one is. Like, ‘here,’ however, sensitivity to certain sources of information
is not enough for ‘I-thoughts.’ There must also be links to action (1982, 209).
Evans presented his results as merely preliminary (1982, 255), but several of his
theories have become widely influential, particularly his account of the self-ascription of
belief. One reason for the popularity of this piece of Evans’s work is that it seems to
provide a model of self-knowledge that
avoid[s] the idea of this kind of self-knowledge as a form of perception—mysterious in being
incapable of delivering false results (1982, 225).
3
Starting from an epigram of Wittgenstein’s he argues that we are led to the Cartesian
position by the assumption that we know our minds through taking an inward glance at
our states and activities. That is the mistake he tries to correct:
In making a self-ascription of belief, one’s eyes are, so to speak, or occasionally literally,
directed outward—upon the world. If someone asks me ‘Do you think there is going to be a
third world war?’, I must attend, in answering him, to precisely the same outward phenomena as
I would attend to if I were answering the question ‘Will there be a third world war?’ I get myself
in a position to answer the question whether I believe that p by putting into operation whatever
procedure I have for answering the question whether p...
We can encapsulate this procedure for answering questions about what one believes in
the following simple rule: whenever you are in a position to assert that p, you are ipso facto in a
position to assert ‘I believe that p’. (1982, 225-26)
Evans goes on to say that other elements must be involved in possessing the concept
of ‘belief,’ such as understanding that the psychological concept ‘ξ believes that p’ can
be instantiated otherwise than by oneself (1982, 226).
Still, the further abilities required for using the predicate ‘believes that p,’ abilities
such as appreciating the kinds of evidence that would lead one to attribute the concept
to another, are background conditions and
adding the background makes no difference to the method of self-ascription: in particular, we
continue to have no need of the inward glance. (1982, 226)
With the background in place all that is required on the occasion of belief self-ascription
is looking outward. Among the many philosophers influenced by Evans Richard Moran
(2001) takes the view that the inner mental world is ‘transparent’4 to the outer world as a
starting place for his account of the authority of self-ascriptions of belief. As we see
below, Kant’s theory implies that for a significant class of beliefs (or judgments), and
perhaps for all of them, the Wittgenstein-Evans project of ascribing self-knowledge
solely through looking outward must fail.
The third current assumption that is undermined by Kant’s work is not usually
highlighted. It is implicit in the discussions of Strawson and Evans, but far more evident
in Elizabeth Anscombe’s seminal discussion of ‘The First Person,’ so I consider it in that
context. Anscombe’s focus is the odd situation of ‘I-thoughts.’ The difficulty is that any
solution to the problem of the reference of ‘I’ must be such that reference is guaranteed
4
(1975, 56). In that case, however, the referent cannot be the body, so it could only be
the Cartesian ego.
Anscombe supports this line of argument via her famous Gedanken-experiment
with a sensory deprivation tank. In the tank Anscombe would be without sight, unable
to speak, unable to touch her body, and locally anaesthetized. Under these
circumstances the object of ‘I’ could not be her body, because that object is in no way
present to her. (Here she makes the Fregean assumptions that the referent of a term is
determined by its sense and that the sense is the way in which the object is presented
to the person [1975, 55]) Yet she thinks that she might well tell herself:
‘I won’t let this happen again.’ (1975, 58)
That is, she believes that she would not have lost her self-consciousness nor what she
means by ‘I.’ In the sensory deprivation tank, however, there is nothing that could be
present to her but a Cartesian ego, viz., ‘the thinking that thinks this thought’ (1975, 58).
Thus we discover that, if ‘I’ is a referring expression, then Descartes was right about what the
referent was. His position has, however, the intolerable difficulty of requiring an identification of
the same referent in different ‘I’-thoughts (1975, 58).
Even though a person in the tank could allegedly attach ‘I’ to a present thinking, she
could never be in a position to attach the same ‘I’ to different thinkings. Anscombe
takes this to be a reductio of the Cartesian approach and concludes that the only way
out is to admit that ‘I’ is not a referring expression at all (1975, 59-60).
This conclusion has not been widely embraced. Evans twice describes it as
‘extraordinary’ (1982, 212n., 214n.). Other philosophers may question her reliance on
controversial Fregean doctrines. Still others may wonder how she thinks reference to
‘this thought’ is possible. Without an answer to that question, it is very unclear how ‘the
thinking that thinks this thought’ could be attached to any referent. The Kantian
objection I press below rests on a different consideration. Anscombe believes that
philosophers should try to find a sense for ‘I’ before turning to the problem of its
referent. Yet she assumes that her readers will agree that she can meaningfully use ‘I’
in the sensory deprivation tank—can use it to refer (1975, 58)—even though the issue
of the identity of the ‘I’ through time has not been addressed. This is the common,
implicit assumption that becomes explicit in the Gedanken-experiment: It is possible to
5
address the referring use of ‘I’ without considering the unity of the thought in question
with other thoughts. From Kant’s perspective, if ‘I’ or ‘I-think’ can legitimately be used,
then the thought must already be understood as united with others in a single self.
To sum up, many contemporary philosophical theories of self-knowledge are
shaped by three assumptions:
1. First person-third person asymmetry: Self-knowledge of mental states such as
beliefs is fundamentally different from knowledge of the beliefs of others.
2. Transparency: Self-knowledge of beliefs can be explained solely through beliefs
about the ‘external’ world.
3. Self-consciousness without unity of self-consciousness: The self-ascription of
belief can be explained (or, in Anscombe’s case, explained away) without
considering the co-belongingness of the belief with others in a common
consciousness.
I take Strawson and Anscombe to hold one and three and Evans to hold all three.5
2. Kant’s Account of the Unity of Self-Consciousness
To see the mismatch between Kant’s views and these contemporary dogmas, we
need to examine his theory of the transcendental unity of self-consciousness (or
apperception). It is a crucial fact about the theory that it is introduced in the context of
explaining the possibility of cognition. He believed that there were different levels of
cognition, some of which were enjoyed by animals. Animals represent objects, perceive
them and differentiate them with respect to sameness and difference (9.64-65).6 But
they are incapable of the rational cognition distinctive of humans—cognition where the
subject knows the basis of her cognition. An ox might, for example, differentiate his stall
because of its door, but he could not know that that is the ground of his cognition (2.5960). The ability to know the ground of a cognition is sophisticated, because it involves
what contemporary theorists call ‘metacognition.’ A person who knows the basis of her
judgment does not just make judgments; she also knows something about them.
6
The cognition at issue in the Critique of Pure Reason is conceptual cognition and
Kant argues, via contrast cases such as that of the ox, that human concept users know
the reason when then apply concepts. At a minimum they know that the correct use of
a concept involves some type of similarity across the objects named by the concept.7 In
the A edition, the unity of apperception is shown to be a requirement for rational
conceptual cognition through an example:
If, in counting, I forget that the units that now float before my mind or senses were added
together by me one after another, I should never cognize the amount being produced through
this successive addition of unit to unit; nor, therefore, would I cognize the number. ...
[T]his one consciousness is what unites in one representation what is manifold, intuited little by
little, and then also reproduced. Often this consciousness may be only faint, so that we do not
[notice it] in the act itself, i.e. do not connect it directly with the representation’s production, but
[notice it] only in the act’s effect. Yet, despite these differences, a consciousness must always
be encountered, even if it lacks striking clarity; without this consciousness, concepts, and along
with them cognition of objects, are quite impossible. (A 103-104, my underscoring)
Kant makes two different points in the passage. The first is that a subject could not
count the members of a set, could not determine its size, if she keeps losing the
information that she acquires about it members: it has at least one member, it has at
least two members, and so forth. Subjects do not lose this information, because they
(obviously) have a faculty of reproductive imagination that enables them to retain and
re-access this information. The second point is the key. It is not enough that this
information is somehow preserved and accessible. Rather, in the act of making a
judgment, say ‘nine,’ the subject is and must be conscious of making the judgment
through combining information in it. He allows that thinkers do not always pay attention
to individual steps, adding up the stroke symbols little by little. Nevertheless, they are
conscious in judging ‘nine’ of doing so through carrying out these steps; if they could not
be so conscious, then they would lack all rational cognition.
Through conscious acts of combining rational cognizers know—as they must—
the bases of their judgments. Rational cognition also involves the unity of apperception
or self-consciousness. Kant takes the concept of a cognitive subject to include the idea
of a subject whose states necessarily belong together (A107, A116, A124, B132, B137).
7
Rejecting as indefensible his predecessors’ claims that the concept is just a special
case of the metaphysical truth that accidents (states) must belong to substances (a
soul), he argues instead that the principle that different mental states are necessarily
connected and the corresponding concept of a cognitive subject reflect a necessary
condition for cognition. Through engaging in rational cognition a subject forges relations
of rational dependency and so necessary connection across mental states. In Kant’s
terms, she ‘brings’ them to or under the unity of apperception (A108, B134-35, B13637). Because she forges these relations, she understands, as she must, that they
obtain. She understands that the state of judging ‘nine’ is rationally and existentially
dependent on other states (though she would not express what she knows in these
philosophical terms.) Failing this recognition, she would not know the basis of the
judgment.
The A edition argues further that the necessary unity of different mental states in a
single subject can be understood only in relation to rational cognition. Empirical
apperception or introspection could never supply evidence for a necessary connection
across mental states (A107). For this reason, Kant argues that the representation ‘Ithink’ must be ‘a priori.’ What that means, roughly, is that humans must come with
capacity to combine representations in such a way that they can thereby form the
representation of a subject whose states are necessarily connected. (I return to this
issue below.) Although he does not draw this corollary, by the same reasoning, insofar
as thinking is an activity that produces relations of necessary connection across states,
humans must come with an a priori concept of thinking, since they could acquire such a
concept neither from outer nor, pace Locke, from inner sense. They must come with
the capacities to combine mental states and recognize them as combined that are
required for rational cognition, and thereby form a representation of thinking.
The B edition stresses the weakness of introspection in relation to the
representation ‘I-think’ in a different way. Kant seems to be criticizing Locke:
The empirical consciousness that accompanies different representation is intrinsically sporadic
and without any reference to the subject’s identity. (B133)
8
His objection is that Lockean consciousness—the consciousness which is inseparable
from thinking (Locke, 1690, 2:27.9)—is momentary or episodic. It’s not just that
empirical evidence could never supply evidence for necessary connection. Lockean
consciousness could exist in the absence of any identity of subject across the states.
By contrast rational cognizing guarantees the necessary connections across states
that are the hallmark of belonging to the same subject. In the B Deduction, as in the A
Deduction, the key is conscious acts of synthesizing.
Hence this reference [to an identical subject] comes about not through my merely
accompanying each representation with consciousness, but through my adding one
representation to another and being conscious of their synthesis. Hence only because I can
combine a manifold of given representations in one consciousness, is it possible for me to
represent the identity of the consciousness itself in these representations. (B133, my
underscoring)
Through engaging in conscious acts of rational cognition a subject forges the relation of
rational dependence and so necessary connection across mental states that make them
states of a single subject. Further, rational cognition requires that subjects recognize
the relation of rational dependence as such (though they would not use these terms).
They understand, for example, that their judgment ‘nine’ depends on their having gone
through the counting procedure.
To press a Kantian critique, I need to show that his analysis of rational cognition
and its connection to the unity of self-consciousness is correct or at least plausible.
That task is made more difficult by what is now obvious: his position is not minimalist.
Pace Strawson, it is not all the same to Kant whether one says that the unity of ‘I-think’
must accompany all my perceptions or none. Rational cognition would be impossible
without a relation of recognized rational dependence across mental states. Further, it
may be analytic that all my states are mine, and that all rational cognitions depend on
reasons. It has not, however, seemed obvious that human cognizers must regard their
mental states as standing in relations of rational and so necessary connection. Nor is
the cognition distinctive of humans widely recognized as requiring conscious acts of
combining or synthesizing. Yet the theory of the transcendental unity of apperception
9
asserts that any case of the self-consciousness distinctive to human cognizers8 involves
the unity of self-consciousness and that the unity of self-consciousness is possible only
through the conscious acts of combining some cognitive states in others that are
essential to rational cognition. In the remainder of this section, I try to show that the
theory is sufficiently plausible to provide a basis of criticizing others, by raising and
rebutting two serious objections. I provide further considerations in its favor when using
it to criticize the three assumptions of contemporary theories of self-knowledge
presented in section 1.
A common aim of contemporary work on self-knowledge is avoiding questionable
mental processes, so going back to Kant’s theory of conscious combining may seem to
be a giant step in the wrong direction. To see that conscious combining is both a
familiar phenomenon and necessary for rational cognition, consider the most obvious
case of cognition where the subject knows the reason, that of inference. For example:
If the economy is bad, then incumbents lose.
The economy is bad.
Therefore, incumbents (will) lose.
Nothing esoteric is going on here. After reading the premises, normal humans draw the
conclusion even before seeing it. And they are conscious in the act of drawing the
conclusion, as they must be, if they are to understand the conclusion as such, as based
on the premises.9 Kant rejected the Lockean hypothesis of inner sense as an adequate
explanation for rational cognition because it involved only the passive receipt of
information and not active combining. Through inner sense subjects would be
conscious of the premises and conscious of the conclusion, but would not consciously
draw the conclusion on the basis of grasping and combining the premises. Hence they
would not know the connection between the premises and conclusion and so the basis
of the judgment.
Although the idea that people consciously perform cognitive acts was widely
accepted in the Early Modern period (see, e.g, the Port Royal Logic, Arnauld
[1662/1964]), mental act awareness has only recently returned to philosophical
10
discussions. When considered it is often limited to unusual cases.10 So, for example,
Alfred Mele offers the example of trying to think of seven animals whose names begin
with G (2009, 17ff.). Tyler Burge argues that to have the first-person concept, one must
be capable of critical reasoning and
to be capable of critical reasoning, and to be subject to certain rational norms necessarily
associated with such reasoning, some mental acts and states must be knowledgeably
reviewable (1996, 98).
Anyone who can use the first person concept and can engage in rational cognition must
also be capable of—occasionally—engaging in second-order thought about reasons
and mental actions. My appeal to the somewhat artificial case of formal inference may
also suggest that mental act awareness is rare.
Kant’s view is that mental act awareness is necessary to rational cognition and
so pervasive in human mental life. Many reasons stand behind his choice of the
counting example,11 but one is presumably its ordinariness. We can get a better sense
of the ubiquity of mental act awareness by looking at two quotidian activities, driving
home from work and shopping for Christmas presents. Presumably most of the ‘other
ninety-nine percent’ of Americans look at the prices of possible presents. Buying
Christmas presents often requires serious amounts of reflection, but my concern is not
with judgments about what to buy, but with judgments about what things cost. If a
shopper judges that the Lego helicopter is $15.95, then he is aware of judging on a
particular basis: he ran the bar code under a scanner, he saw an ad in the paper, he
saw the sign above the helicopter boxes, or he used some other source of information.
The mental movement may be less noticeable than in the case of inference, but he is
still aware of making the judgment through, e.g., seeing the sign. This is still a case of
rational cognition. Now consider an example that arises in some discussions of
consciousness: driving home while your attention is completely absorbed in thinking
about something else. While not rare, this experience can be unnerving, since it is
puzzling how such a complicated task could be accomplished ‘unconsciously.’
My point is that the conscious acts of judging involved in Christmas shopping (or
in watching a movie or a baseball game) are not unusual. They are not recherché or
11
reflective states, but the stuff of everyday rational cognition. We sometimes lose track
of what we are doing either through having our attention focused on thinking about other
matters, as in the distracted driving case, or simply by day-dreaming, but for most
people such lapses are not common. We go about our lives aware of what we are
doing, including being aware of acquiring information needed for our projects from
various sources and then using that information to determine, e.g., that the Lego
helicopter is less expensive than the Storm Trooper helmet.12 Wittgenstein mocks a
position very like Kant’s when he asks rhetorically, ‘did you recognize the desk in your
office?’ All agree that you do not judge that there is a desk in your office every time you
enter it, since you already have the information (1953/1997, 602-604, p. 157). But this
doesn’t touch Kant’s point that you made that judgment consciously the first time.
My defense of Kant’s appeal to the mental act awareness involved in rational
cognition has been limited: It is not a philosophical invention, but a widespread
phenomenon that is easily recognized once attention is drawn to it. I have not provided
a theory of this awareness, but treated it, as Kant does, as part of the explication of
rational cognition, rather than as an explanandum for theories of mind. In using his
theory to criticize others below, I bring out some aspects of mental act awareness, but
these observations fall short of a full account. I will, however, address an obvious
objection from the inner sense debates. Locke’s hypothesis of an ‘internal sense’ is
usually regarded as incoherent, because the sense is supposed to be both perceptual
and infallible. Kant’s theory cannot be criticized on exactly these grounds, because
transcendental apperception is not a kind of perception, but a kind of action. Still, to lay
the ghost of a magical kind of consciousness to rest, I clarify the role of mental act
awareness in self-knowledge of beliefs or judgments.
Tyler Burge has recently argued that the history of epistemology is plagued by a
failure to distinguish between philosophical theories of the necessary conditions for
psychological states to have objective reference and the psychological capacities
involved in possessing the states (2010, 12-22). Burge’s focus is perception and he
takes Kant’s to be the demanding sort of cognition required for science (2010, 155, n. 4,
156). As is evident in my examples, I think Kantian rational cognition has a far wider
12
scope, but I won’t address this long-standing interpretive issue.13 The Critique of Pure
Reason is among Burge’s targets only because he thinks that it inspired philosophers
such as Strawson and Evans to conflate accounts of how subjects understand
objectivity or objective reference with accounts of objective reference per se. Still
Burge’s fundamental complaint, that these theories ‘hyper-intellectualize’ (2010, 12),
can seem tempting in relation to Kant. In the present context, the charge would be that
his claim that rational cognizers must possess the unity of apperception—they must
forge and recognize relations of rational dependence across their states—is an
intellectualist myth. Perhaps philosophers can show that the concept of rational
cognition presupposes the concept of a unified subject. That philosophical theory
should not, however, be offered as an account of the psychological processes or
capacities of rational cognizers. Making this move would be just the sort of ‘armchair
argument’ about the relations among psychological abilities that Burge decries (2010,
209).
Kant’s theoretical interests were not psychological, but philosophical. Via the
counting example, he investigates what the phenomenon of rational cognition involves:
combining thoughts in a resultant thought that is understood as having been produced
by the combining. He then considers the relation between rational cognizing and the
philosophical problem of the unity of the mind. In rational cognizing subjects combine
materials from different states in another state thereby making a connection across the
states that they recognize as such. The phenomenon of rational cognition thus
necessarily involves the unity of self-consciousness. More explicitly, Kant starts with
the fact that humans have a capacity for rational cognition, a capacity that involves
producing and recognizing a relation of rational dependence across their states. He
argues that rational cognition must involve both the production and the recognition of
this relation by raising the possibility of a case where there is no act-consciousness
(and hence no recognition of dependency [A103-104, cited above]) and noting that it
would fail to be rational cognition. He (implicitly) notes that relations of rational
dependency are relations of necessary connection. The analysis of rational cognition
and of the connection between rational dependence and necessary connection leads to
13
the philosophical claim that it is necessary for rational cognition that different mental
states stand in relations of necessary connection and are recognized by cognizers as
standing in relations that philosophers describe as relations of necessary connection.
Kant’s theory of apperception is minimal in one respect: He does not have a
metaphysically laden account of a unified consciousness, but understands the ‘unity of
self-consciousness’ just to indicate relations of necessary connection across mental
states. Given this minimal conception of the unity of consciousness, he can move from
the preceding philosophical claim to his philosophical thesis that the unity of selfconsciousness or apperception is necessary for rational cognition. Alternatively, when
he presents the faculty of conscious combining as a special faculty—‘transcendental
apperception’ (A94-95, A106-106, B132, B134n.)—he is not hypothesizing a novel
capacity, but highlighting the philosophical importance of his analysis of rational
cognition. His analysis of what rational cognition involves shows that it requires a
faculty that consciously combines some states in others and recognizes the relation of
dependence thus produced, the relation that is the hallmark of belonging to a common
subject. It requires a faculty that can be perspicuously labeled ‘transcendental
(necessary for cognition) apperception’ (self-consciousness, because the states
consciously combined are states of the subject) and that produces what can
perspicuously be labeled the transcendental (necessary for cognition) unity of
apperception (the recognized relation of necessary connection across the states of the
subject that have been combined).
Although Kant’s investigation of the philosophical implications of rational
cognition does not posit psychological capacities, it depends on and bears on cognitive
psychology. Future research could up-end his philosophical results by showing that
‘rational’ cognition is not as it seems. Perhaps what people do in ‘giving the reason’ is
just say out loud whatever thought they had immediately preceding the judgment.
Perhaps if asked they would deny that the judgment depends on other thoughts. Some
two-year olds can count in the sense that, if asked, they can correctly give the
cardinality of a set of proffered objects: 1, 2, 3, 4, 5. Others can separate the cardinality
from the procedure, by repeating the value when they finish: 1,2,3,4,5 …5. Even these
14
more adept counters don’t seem to understand what they have done. When asked
repeatedly to give the number for the same set of objects, they simply repeat the
procedure: 1,2,3,4,5 …5. By contrast, 3-year-olds just give the answer: 5 (KarmiloffSmith, 1992, 102). If all apparently rational cognition were like the counting of 2-year
olds—if it involved no metacognition of the production of the judgment—then Kant’s
appeal to it to solve the philosophical problem of the unity of the ‘I-think’ would
collapse.14
Although humans appear to have a capacity for rational cognition that involves
metacognition, my defense against the charge of hyper-intellectualizing is not that
laypeople are ‘lay epistemologists.’ Sometimes developmental psychologists describe
children as ‘little linguists’ or ‘little scientists.’ The temptation to do so is evident when
presented with data such as the following:
‘What’s that?’
(Mother: ‘A typewriter’)
‘No, you’re the typewriter, that is a typewrite.’
(Yara, 4 [cited in Karmiloff-Smith, 1992, 32])
What this exchange shows is that even a 4-year old has metalinguistic knowledge: She
has some appreciation of the regular way in which English marks agency. It does not
show that she understands what it would be to make language an object of study. Kant
did not think that ordinary people understood what it was to make knowledge a subject
of study;15 he thought that they possessed sophisticated cognitive capacities that
involved what we would call metacognition. As noted, however, although the capacity is
complex, it is not uncommon. People know the reason when they judge that it is safe to
cross the street, that the restaurant is full or that the baby is hungry.
Kant’s results can offer guidance to cognitive psychology. He offers partial
accounts of what is involved in rational cognition and in having a unified selfconsciousness thereby establishing a connection between the two capacities. He also
highlights the explanatory inadequacy of then current theories of the psychological
faculty that supports these capacities, viz., inner sense. Burge accepts Marr’s schema
15
for the construction of complete accounts of psychological capacities, which includes a
high-level description of a capacity (2010, xvii-xviii), so he might not object to Kant’s
analyses providing direction and constraints for fuller accounts of important
psychological capacities.16 Far from offering a psychological hypothesis about how
‘transcendental apperception’ might work, Kant found the capacity baffling. When trying
to characterize it explicitly, to distinguish it from inner sense, he settled on a negative
characterization. The synthesis produced through transcendental apperception17
if the synthesis is viewed by itself alone, is nothing but the unity of the act, of which as an act, it
is conscious to itself even without sensibility. (B153)
I have tried to turn away the obvious objections that Kant’s theory of the
transcendental unity of apperception credits laymen who are capable of cognizing with a
philosophical theory of cognition and that it does so by invoking a conscious—yet
completely unsuspected—faculty. These defenses are not intended as final and we’ll
return to the issues in weighing Kantian criticisms of other approaches. We’ll also see
some shortcomings of his views, those these do not, I argue, undermine the critiques.
3. Kant’s Theory and Anscombe’s Experiment
In ‘The First Person’, Anscombe argues that ‘I’ cannot be a referring expression,
because if it were, the only possible referent would be a Cartesian ego, i.e., the thinking
that thinks this thought (1975, 58), and such a referent cannot have temporal duration.
In particular, in a sensory deprivation tank, there is no way to identify the thinking that
thinks this thought at t1, the thinking that thinks this thought at t2 and so forth. If Kant is
right that engaging in rational cognition requires forging and recognizing the rational and
necessary connections across different states, then Anscombe’s thought-experiment is
misdescribed. Since her exclamation ‘I won’t let this happen again’ is a case of rational
cognition, she must tacitly understand its dependence on other states, specifically, on
the very odd state of having no current sensory stimulation, on her recollections of
previous sensory states and so forth. She could not be in the situation where she refers
to the thinking that thinks only this (current) thought, ‘I won’t let this happen again.’ Nor
16
is it accidental that Anscombe chooses a case of rational cognition. Had she opted for
something more primitive, e.g., ‘I’ followed by a groan, then questions would arise about
whether the producer of the sound ‘I’ was an ‘I.’
On one point, Kant agrees with Anscombe: It is impossible to identify the same
Cartesian ego, understood as ‘the thinking that thinks this thought,’ across different
thoughts. Such ‘thinkings’ would be episodic and have no reference to a common
subject. This was just his criticism of (probably) Locke in the B Deduction. He would
object, however, that Anscombe presents a false picture in suggesting that there can be
‘thinking’ and ‘I’s’ whatever the circumstances. In his view, thinkers are not born (met
lectures), but ‘self-created’ through exercising their powers of combination on materials
that are combinable by them ultimately in judgments, through bringing different
representations to the unity of apperception. That is why it is impossible, in principle, to
present a legitimate use of ‘I’ while still leaving open the question of the unity of the ‘I’ or
‘I-think.’ If rational cognizing is going on—if there is a thinker—then different thoughts
belong, and must be recognized as belonging, to a common thinker. Cognizers identify
the I-think across different thoughts because they combine thoughts, because they
think.
Further, Kantian subjects are not the ‘short’ selves that Anscombe mentions and
dismisses (1975, 58). A subject who consciously judges A at t1 and B at t2 could always
judge ‘A and B’ at t3, thereby making the co-belongingness of the states evident. As
often noted, Kant’s claim is that states belong to the same subject if it is possible to
attribute them to a common thinker (B132). Since it is always possible for Kantian
subjects to connect a judgment to a previous or a subsequent one, they could form
chains of indeterminate length.18
A defender of Anscombe could argue that she offers the characterization ‘the
thinking that thinks this thought’ in order to capture the fact that the use of ‘I’ cannot be
mistaken. It cannot fail to refer and it cannot refer incorrectly (1975, 56-57). Kant has
not explained how ‘I’s infallibly refer to themselves—or even how they refer to
themselves at all. Even if a rational cognizer must recognize different states as
17
rationally interdependent and so as belonging to a common subject, how, on Kant’s
view, is she justified in attaching ‘I-think’ to the representations?
Kant has an explicit and incorrect answer to this question: The thinker is aware of
the states through inner sense (A22/B37). Since he relies on inner sense to distinguish
a subject’s mental life from the rest of the world, he does not address the question of a
criterionless and so infallible referring use of ‘I.’19 Still, a Kantian account of infallible
self-reference falls out of his theory of rational cognition. We can begin with some
standard assumptions. The referent of ‘I’ is picked out by the self-reference or selfreflexive rule: ‘I’ refers to the ‘thinker of this thought’ (see, e.g., Peacocke, 2008,103).
Given Kant’s emphasis on the active character of thought, another way to express the
self-reflexive rule might be ‘“I” refers to the author or maker of this thought.’20 Subjects
would be able to use ‘I’ by learning the self-reference rule. They know to say ‘I think’
when they are thinking. But how do they know when they are thinking?
It is at the point that distinctively Kantian elements enter the picture. One could
protest that no further explication is needed, because the correct answer to this
question is either ‘they just do’ or ‘leave that for the psychologists.’ As noted, however,
Kant’s analysis of rational cognition provides guidance for cognitive psychology. On his
account, what thinking qua rational cognition amounts to is that a subject consciously
combines thoughts in a further thought thereby creating and recognizing a relation of
necessary connection across them. Subjects know that they are acting because they
consciously combine and so are conscious of combining; they know they are acting in a
particular manner, thinking, because they recognize the relations of rational
dependence and necessary connection that they have forged across their mental
states.
In order to avoid transparent circularity, these conditions need to be expressed
carefully: subjects are conscious of combining—not conscious that they are
combining—and they are conscious of a relation across states that are theirs, not
across states that they take to be theirs. Even though a subject is not conscious that he
is combining, neither is he conscious that someone is combining or that combining
going on. He consciously combines and is conscious of combining.21 Since he has an
18
a priori concept of thinking as activity that results in a relation of necessary connection
across thoughts/states, he is aware of what are in fact, his acts, as acts of thinking and
so can apply the self-reflexive rule that ‘I’ is used to refer to the thinker or author of a
thought and say ‘I-think.’
Conscious combining plays a crucial role in enabling cognizers to apply the selfreference rule, because that is how they know that they meet the condition specified in
the rule. They are conscious in thinking and so conscious of thinking. But this
consciousness is importantly different from the consciousness allegedly provided
through inner sense. The latter was supposed to enable subjects directly to perceive
their thoughts. A subject who consciously combines some thoughts in another that she
thereby recognizes as rationally dependent on those thoughts cannot be mistaken that
she is judging, not because the consciousness is a magical kind of perception, but
because that is what judging is. In this way, Kant’s account of what rational cognition
involves provides an account of how and why self-ascription of judgments is infallible:
The infallibility of uses of ‘I-think’ is a product of both the unusual character of the selfreference rule and the unusual character of thinking. Subjects who consciously
combine some thoughts in others and thereby recognize the relations of rational
dependency across the thoughts do not merely think that they are thinking; they are
thinking.22
The latter part of this section argues against Anscombe’s conclusion that ‘I’
cannot be a referring expression at all, because it is impossible to account for both the
infallibility of its reference and the temporal duration of its referents. It was offered to
counter the objection that Kant’s theory is too defective to be used as a basis for
criticizing other theories of self-knowledge. Although he was mistaken in relying on
inner sense, this piece of the theory can be replaced with something suitable to the
task. My target is not, however, Anscombe’s unpopular conclusion that ‘I’ does not
refer, but her oft-cited sensory deprivation tank thought experiment. If Kant is right, then
the description of the case is inconsistent. She envisions a situation where she can
self-ascribe a rational cognition without first understanding its relation to other thoughts
in a single subject. Many discussions introduce the topics of self-knowledge or self19
consciousness via the three striking features of infallibility, authority, and immediacy. In
so doing, they assume that these issues can be resolved without first (or possibly ever)
considering the unity of consciousness. Kant’s theory of the unity of apperception
implies that this assumption is mistaken. Further, on the ‘Kantian’ account of the
infallible self-ascription of judgments—that judging is a matter of consciously combining
some states in others and of recognizing that combination as such—subjects who are
capable of infallible self-ascription of judgments must also be aware of the unity of
different states in a single subject.
4. Implications of Kantian Unity for Transparency
Kant’s account of rational cognition has equally direct implications for the
‘transparency’ approach, the assumption that knowledge of the ‘inner’ is grounded
solely in knowledge of the ‘outer.’ The implications are evident as soon as we consider
a possible reply to Evans’s enquiry about a third world war. Suppose that Frank
reasons as follows:
Wars are brought about by the preparation for war.
Major powers are now disarming.
Therefore, a third world war is unlikely.
Frank forms a judgment or belief by looking outward, by learning or recalling the
correlation between war and levels of armaments, by considering the current level of
arms. He cannot, however, just be looking outward. He does not ‘look’ inward, but if
the judgment is rational, then he must know that he makes it on the basis of something
else, in this case, the correlation and the current situation and the realization of what
they imply. Further, he does not know the basis by reflecting on why he thinks a third
world war is unlikely and canvassing some likely explanations. This is not an exercise
in critical reasoning (Burge, 1996, 98-99). Like the Christmas shopper who determines
the price, he forms the judgment that a third world war is unlikely and knows the basis of
that judgment in a single mental act. Thus, to judge (or to come to believe) that a third
world war is unlikely, Frank must know things about the mental as well as the external
20
world: The thought that a third world war is unlikely did not arrive as an inspiration from
the gods. It was produced by reasoning from particular facts or thoughts.
In section 2 I tried to rebut the general objection that Kant confused a philosophical
theory of cognition with an account of the psychological capacities required for
cognition. It may be tempting to try to defend transparency by raising a special case of
that objection: Although epistemologists and psychologists know that mental processing
is required to form rational judgments, cognizers can form them without taking account
of any mental conditions. All that Frank needs to form the belief that a third world war is
unlikely is knowledge of the correlation and of the facts and the ability to infer the
judgment from them.23 Since the correlation, the facts, and the logical relations are
independent of Frank’s mind, he can form the belief just by looking ‘outward.’
For reasons that will be obvious, I discuss this objection in relation to inference, but
the same considerations can be raised for any judgments/beliefs where the subject
knows his grounds. Since the capacity is rational cognition, Frank’s psychological
processes cannot involve just the inputting of the premises and the outputting of the
conclusion. That could happen unconsciously and he would have no idea why he is
claiming that a third world war is unlikely. This would not be a case of Kantian rational
cognition or of belief, as Evans seems to understand it.24 Suppose that we add that
Frank is conscious of the premises and conscious of the conclusion, and that his mind
moves from the premises to the conclusion—but that he is not conscious in drawing the
conclusion. Lewis Carroll famously argued that reasoning requires not just premises
and inference rules, but actual inferring, mental movement from premises to conclusion
(1895).25 On Kant’s analysis, that is still insufficient. Taking up the role of the Tortoise,
Frank could say that he is conscious of the premises and that, when he is conscious of
the premises, he quickly becomes conscious of the conclusion (because his mind does
move), yet deny that he believes that a third world war unlikely because of the
correlation and the facts—because his mental action is unconscious. That position
seems as unassailable as the original Tortoise’s claim that he accepts the premises of
an instance of modus ponens but not the conclusion. If Frank is conscious not just of
the premises and of the conclusion, but also in drawing the conclusion on the basis of
21
the premises, however, then he makes his judgment using mental as well as external
factors, because he understands the judgment as produced by thinking from other
thoughts and so recognizes its rational and existential dependence on those thoughts
(though he wouldn’t express the point in these terms).
A transparency advocate might respond that the Kantian description of Frank’s
situation is tendentious, even question-begging. Why is it not enough to say that since
Frank is conscious in drawing the conclusion on the basis of the premises, he sees its
dependence not on particular thoughts, but on particular facts? The objection can be
cast in terms of a dilemma: Either the Kantian holds that Frank is conscious of the
dependence of his judgment about the third world war on his beliefs about the
correlation and the current state of disarming, in which case the account is circular, or
he holds that Frank is conscious of the dependence of the judgment on the facts about
the correlation and disarmament, in which case the transparency strategy has not been
undermined. I counter this defense, first by looking at how the Kantian account avoids
circularity from a slightly different angle and second by offering a further consideration
that seems to force the ‘thought’ rather than ‘fact’ account of rational cognition.
The circularity problem can also be presented in terms of an infinite regress. If
Frank believes that his beliefs in the correlation and in the facts produce his belief that
there will likely be no third world war, then how did he self-ascribe the beliefs that are
the reasons for the target belief? We can see more clearly how the Kantian position
avoids circularity, by recalling his view that thinkers are not born, but self-made through
thinking. Absent thinking, there is no thinker to whom a thought could be self-ascribed.
Nor is there a thought that could be self-ascribed. Suppose that Frank learns of the
correlation, recalls the efforts at disarmament and concludes that a third world war is
unlikely. He need not pre-identify the thoughts that constitute the premises as ‘his;’ he
need not even pre-identify them as ‘thoughts.’ He simply has various mental states,
consciously draws the conclusion from them, thereby forging and recognizing the
relation of rational dependence across them that make them thoughts of a rational
cognizer. Circularity is avoided on Kant’s position in the same way that it is avoided on
Evans’s, because it is not the belief that you have a thought that produces other
22
thoughts, but engaging in thinking that leads to the thought that is the conclusion—and
which creates the required relation of recognized necessary connection across the
thoughts that enables the thinker to recognize that he is thinking and to attach the ‘I’ via
the self-reference rule.26 On Kant’s view, but not on Evans’s, engaging in rational
cognizing produces not just a judgment, but also an understanding of that judgment’s
production through the mental manipulation of what are thereby understood as
‘thoughts.’
We can see why rational cognition should be described in terms of seeing a relation
across thoughts, rather than across facts, by turning to an issue that I have so far
avoided. Why must thinking involve mental actions? Henry Allison has argued that, for
Kant, thinking is acting, because it is free (1990, 36, 2006, 389). I don’t think this is the
way to understand Kant,27 and the examples already given indicate why it would be an
implausible account of thinking. Humans are not ‘free’ to refrain from making simple
inferences any more than they are free not to hear their own names when they are
spoken. It seems equally implausible to claim that thinking must be understood as
acting because it is produced by intentions. Troubled souls may intend to block thinking
through drink or drugs, but their more fortunate brethren do not form intentions to think.
So how can Kant be right that thinking or rational cognizing necessarily involves a type
of action?
Kant originally presents thinking as acting by contrasting it with sensibility, which
involves the passive receipt of information rather than its active combination (A51/B75,
A68/B93, B129-30). Insofar as the distinction is just that between ‘passive and active
powers,’ however, thinking would no more be a case of acting than digesting is. As we
have seen, he also takes the synthesizing or combining that produces conceptual
cognition to be conscious. When contrasting ‘inner sense’ and ‘transcendental
apperception,’ he stresses that the power of apperception or understanding is conscious
of its acts, even apart from sensibility (B153). He denies that consciousness is
necessary for action (B130), because he believes that cognition requires a great deal of
unconscious processing or combining. Still, the combinings at issue in rational
23
cognition are conscious and that is probably the best way to justify his claim that they
are acts.
Ironically, the classic reasoning linking actions to consciousness was supplied by
Anscombe in her ground-breaking work, Intention (1957/1979). She argues at great
length that action must be intentional and that it is marked by its connection to a special
sort of refusal. A person denies that she performed an intentional action by saying ‘I
didn’t know I was doing that.’ Conversely, if a person cannot deny that she knew she
what she was doing, then she cannot deny that she was engaging in intentional action.
Since rational cognition must be conscious, thinkers cannot deny knowing what they are
thinking. Hence rational cognition has the hallmark of intentional action. Anscombe
argues further that subjects have knowledge of their intentional actions without
observation (e.g., 1957/1979, 14). As noted, Kant maintains that the action awareness
involved in thought must be non-perceptual. Given Anscombe’s analysis of what
(intentional) action involves, Kant claim that rational cognition must involve (intentional)
action would be correct.
If thinking is (intentional) acting, because the subject (non-perceptually) knows what
she is doing, then what does she think that she is doing? The answer must be that she
thinks that she is changing her mind. If Frank knows what he is doing, then he knows
that in judging that a third world war is unlikely he is not walking or singing—or changing
the course of history. He is changing how he thinks about the likely course of history.
Since Evans’s case is a matter of setting out to change or to make up one’s mind, it
might seem that the thinking should be characterized as (intentional) acting, because it
is intended. We can dispel this appearance by returning to the example of Christmas
shopping. The shopper might hear an announcement that today only Storm Trooper
helmets are $10.95 and thus change his mind about their price, even though he lacked
any intention to do so. His thinking would still be (intentional) action, because he knows
what he is doing, not altering, e.g., the price of the helmets, but what he thinks about
their price. Insofar as Kant is correct that rational cognizing is a kind of (intentional)
action, however, then subjects who think must understand themselves to be thinking. It
is not enough for the cognizer to think; he must know that he is thinking, that the
24
judgment is a product of thinking, and that what are manipulated by thinking are
thoughts. This is not a matter of pre-identifying mental states as thoughts before
thinking, but of understanding oneself to be thinking when one is. In that case,
however, rational cognizing must involve metacognition. And the subject who judges
does not acquire infallible self-knowledge of the mental solely on the basis of his
relations to the ‘external’ world, even his mental relations—his grasp of facts or of
logical relations. Judging involves a mental appreciation of the mental—that the
judgment is a product of thinking and that it rationally depends on other thoughts—so
cognizers cannot acquire self-knowledge just by looking outward.
Evans tries to fill in an important piece of his theory of reference by offering a
preliminary sketch of the reference of ‘I’ that addresses the striking fact that it is
infallible. For the case of believing or judging, he offers a three-part account. The selfascription of belief involves a procedure for forming beliefs about the external world, a
rule for using ‘I’ that explains the peculiar fact that self-reference in the case of belief
does not admit of reference failure, and additional knowledge about the conditions
under which beliefs can be ascribed to other subjects. He sets up the problem of
explaining the infallible self-reference of claims about belief in terms of how a subject
answers the question of what he believes (1982, 224-25), so it has been natural to read
him as offering an account not just of reference but also of self-knowledge. The Kantian
critique of this approach not only explains why it is a mistake to think that selfknowledge can be explained without explicit reference to mental conditions, but also
why the transparency strategy seems so plausible: The ‘move’ from (rational cognition)
‘p’ to ‘I think that p’ is foolproof, because there is no move.28 A cognizer who thinks that
p must already understand the act that she, in fact, performs to be an act of rational
cognition, to be a case of thinking. All that she needs to assert ‘I think that p’ is the selfreference rule for ‘I.’
The Kantian critique of transparency is similar to its critique of Anscombe’s attempt
to explain self-consciousness prior to considering the unity of the self. Evans and his
many followers hope to explain self-knowledge, self-consciousness and self-reference
without first addressing the question of what is involved in being a cognizing self. The
25
lack of focus on the necessary conditions for being a rational cognizer per se is
unsurprising, since a primary goal of Evans’s work is to repudiate and replace the
Cartesian conception of humans as first and foremost thinkers. He offers a rule that
bypasses that project: Whenever you are in a position to assert that p (whatever that
might involve), you are in a position to assert ‘I believe that p’. As a result, he doesn’t
see that rational cognition already involves key elements of self-knowledge, in
particular, knowing of your states that they are rationally connected mental states and
knowing of your actions that they are thinking kinds of actions, and thus cannot provide
a reductive account of it.
Evans’s defenders could, however, turn the tables and complain that Kant’s focus on
what he took to be the sort of cognition distinctive of humans led him to the false
position that humans are only or primarily cognitive subjects. This weakness in his
account can be brought out by a skeptical challenge of the sort that Evans uses to
motivate his dispositional account. Suppose that one body counted from one to five,
then passed its thoughts along to another that completed the count to nine?29 It seems
reasonable to hold that the judgment ‘I think that the number is nine’ would be false. To
avoid such cases, the Kantian account would need to be patched up by something like
Evans’s stipulation that the subject acquires information only in relevant ways and that
additional condition may involve reference to particular bodies (1982, 216). This
criticism seems fair and it turns on Evans’s point that Early Modern accounts of subjects
are too thin. Even if what is most distinctive about humans is their ability to think that
does not mean that many other factors aren’t involved in self-knowledge and selfidentity.30 Still, the skeptical challenge to Kant does not defuse the Kantian critique that
Evans’s account of self-knowledge of belief is itself too thin.
Beyond those already noted, Evans had a further goal in proposing his accounts of
self-reference. He wanted to avoid Idealism, views that collapse the gap between
evidence for a conclusion and the conclusion (1982, 210, n.9). In particular, he wanted
to avoid Idealism about the self. He took selves to be part of an objective order that
should be understood as such (1982, 256). Kant’s theory of apperception is not
vulnerable to one of Evans’s objections to Idealist theories, because there is neither a
26
gap between conscious combining and judging nor a verificationist identification of the
two. Conscious combining is not evidence for judging, but the action that produces
judgment properly so called, the action that makes it true that there is a judgment.
Further, there is a clear sense in which Kantian subjects are parts of the ‘objective
order.’ With respect to objects that are not also subjects, their status as parts of the
objective order is independent of what anyone thinks of them or whether anyone thinks
of them. As noted, on Kant’s view, thinking subjects are created through thinking. But
granting that thinking is needed for there to be thinking subjects does not lead to
Idealism: Actual thinking or believing is at issue, not ‘merely’ a subject’s conception of
thinking or believing or judging. On Kant’s analysis, however, rational cognition also
involves conceiving of thinking as such and of thoughts as necessarily connected.31 It is
essential to rational cognition not just that epistemologists can recognize the states of a
subject as necessarily connected, but that subjects themselves also recognize this
relation (even though they wouldn’t express the relation in these terms). Thus, his view
involves ‘subjectivity’ not merely in the sense of ‘about the subject,’ but in the sense of
‘dependent upon the subject’s conception.’ Still, that does not seem to be a vicious
from of Idealism. Kant’s view does not imply that the subject is somehow beyond the
objective world,32 but rather that a complete theory of the items in the objective world
will need to include an account of the basic psychological capacities required for
subjects to grasp not just logical relations among contents, but also relations of
dependence among the vehicles of content. Another way to express the Kantian
objection to transparency is that it falsely assumes that self-knowledge can be
explained without tackling these issues.
5. Kant’s Argument against First Person-Third Person Asymmetry
Although the expression ‘standard view’ is both over-used and misused, the
Strawsonian thesis that it is a fundamental fact about psychological predicates that they
have different application conditions for the first-person and second/third person cases
should be characterized in these terms. As noted, the doctrine was the backbone of the
27
famous argument of ‘Persons’; it is honored in countless subsequent discussions,
including Evans’s (1982, 226). Strawson’s doctrine can be resisted without going back
to Kant. Sebastian Roedl lays out a clear prima facie argument against it:
It is often said that certain concepts, e.g., concepts of experience and action concepts, are selfascribed in a special manner, different from the way in which they are ascribed to someone
else. Equally frequently is it added that, yet, the same concepts are ascribed in these different
ways. It is not obvious how both can be true. If ways of predicating what seems to be one
concept are simply different, then we must conclude that the corresponding expression is
ambiguous. It does not help to insist that the expression must bear a uniform sense in its first
person and its second person use. (2007,165).
Although Roedl’s criticism of Strawson’s reasoning seems right, it is still useful to look at
Kant’s theories of rational cognition and the unity of apperception, because they
illuminate the error that leads to the apparent ambiguity, viz., the assumption of first
person/third person asymmetry.
Kant’s view that thinking is acting may seem, at first, to offer strong support for the
asymmetry position. Since subjects know that they are thinking by performing acts of
conscious combination, their action-based knowledge of their own minds must be
different in kind from their knowledge of other minds. After all, they cannot think anyone
else’s thoughts. In this case, we do not need to try to work out a Kantian position on
knowledge of ‘other minds,’ because he drew out the implications of his reflections on ‘Ithink’ for the issue as he began the Paralogisms chapter. He intends to use his positive
discussion of the ‘I-think’ as a basis for criticizing the claims of Rational Psychology,33
so he prefaces his discussion (in both editions) with an explanation of how he is able to
do so:
It must, however, seem strange at the very outset that the condition under which I think at all,
and which is therefore merely a characteristic of myself as subject, is to be valid also for
everything that thinks; and that upon a proposition that seems empirical we can presume to
base an apodeictic and universal judgment, viz: that everything that thinks is of such a character
as the pronouncement of self-consciousness asserts of me. The cause of this, however, lies in
the fact that we must necessarily ascribe to things a priori all of the properties that make up the
conditions under which alone we think them (A346/B404-405, my underscoring).
28
It may be difficult to understand Kant’s worry from my presentation of his view, which
stresses his analysis of the necessary conditions for rational cognition. If rational
cognition requires the unity of apperception, then all thinkers must have such a unity.
To see the source of his concern, consider the interpretive question of whether the
Critique assumes that there is rational cognition (there are rational cognizers) or argues
for the ‘fact of cognition.’ The correct answer is probably that it does neither. Instead,
in the key passages where he argues for a connection between rational cognition and
apperception, Kant adopts Descartes’ approach and (tacitly) invites his readers to work
through, e.g., the counting example with him. By doing so, they come to an explicit
understanding that they are rational cognizers, that they know the reasons for their
judgments, and that they know the reasons, because they make the judgment on the
basis of the reasons, and that, because they base the judgment on the reasons, their
thoughts stand in relations of rational (and so necessary) connection.
In going through this exercise with Kant, each reader can establish an existential
proposition: ‘I am a rational cognizer.’ To someone in the grips of Lockean Empiricism,
the proposition can also ‘seem empirical.’34 I perceive my thinking through inner sense.
When the inner sense error is compounded by the view that all generalizations rest on
induction, we arrive at the infamous induction from a single case: All minds are just like
the mind I perceive through inner sense. The argument of ‘Persons’ avoids that dire
consequence through the hypothesis that psychological predicates must be applied to
others through logically adequate behavioral criteria. Given such criteria, knowledge of
other minds would have a suitably wide evidential base.
It is just this Empiricist misunderstanding of knowledge of minds that Kant tries to
forestall in setting the stage for his criticisms of Rational Psychology. When readers
work through the counting example, they do not watch themselves think;35 they engage
in rational cognizing. Because they have the ability consciously to combine some
thoughts in others and thereby to grasp the resulting thought and its progenitors as
rationally and necessarily connected, they can rationally cognize and can understand
what rational cognition involves: It is a matter of basing judgments on thoughts to which
they are thereby understood as necessarily connected. Alternatively, what subjects
29
come to understand through engaging in rational thought is not just how they think, but
how rational cognizing works and, hence, how any rational cognizer must think. They
apply that understanding to others and thus take everything that thinks to have the
same character as their thinking. Since humans understand what it is to think, what it is
to have rational thoughts, through their own model, it follows that in the world they can
understand, all representations involved in rational cognition must belong with others to
an ‘I-think’ (B132).
The procedure of using oneself as a model is justified because the only access that
humans have to the nature of thinking is through thinking. They cannot perceive
thinking through outer or inner sense (B419-20, A546-47/B574-75). It is only by thinking
that they can understand thinking and so what it is for anyone to be a thinker. For this
reason, they can apply the predicates ‘believe,’ ‘judge, ‘think,’ etc., to others only by
using themselves as the model. Far from having divergent application conditions in the
first and third person cases, the application conditions are and must be exactly the
same.
Strawsonians might counter this argument with the obvious point that there is a still
glaring difference between the first and third person cases: I know my thoughts just by
thinking them, whereas I have to divine someone else’s thoughts from what she says
and does. Although this is true, the objection does not address Kant’s position at the
level at which it functions. Neither Kant nor Strawson was interested in particular
knowledge claims. They tried to sketch the general conditions against which particular
claims to knowledge were possible. I might learn the contents of a particular thought
through Frank mentioning that he thinks a third world war is unlikely or by receiving
Morse code from Alpha Centauri. Kant’s point is that I can only come to think, e.g., that
Frank believes that a third world war is unlikely, by using my rational cognizing as a
model for his. That is the necessary general condition that spans the endless variety of
particular knowledge claims about minds.
Stawsonians could also try a different tack. Since the attribution of thoughts to
others must be based on some evidence, his general thesis about the necessary coinstantiation of both mental and physical predicates in persons still seems to hold. A
30
disembodied mind could not be the object of mental attribution. As the Alpha Centauri
example indicates this claim may or may not be true. It would depend on whether nonbodies could send signals. That question is, however, irrelevant to the present issue.
The real lesson of the Alpha Centauri example is that there is no way to delimit the
range of evidence that might prompt a person to believe that she is dealing with a mind.
In that case, however, insofar as ‘application conditions’ are understood as ‘evidential
conditions,’ the application conditions for mental predicates to other minds are irrelevant
to the competent use of mental predicates.36 Insofar as ‘application conditions’ are
understood as the conditions that enable humans to have knowledge of minds, by
Kant’s argument, those conditions must be identical for all cases: first, second and third
person.
Rather than shoring up Strawson’s position, defenders could argue that Kant’s
argument for symmetry rests on dubious assumptions. In particular, they might argue
that in maintaining that thinking cannot be represented by predicates that are applied
through ‘outer sense’ (A22/B37, B419-20), he simply begs the question against the
Wittgensteinean mantra that the human body is the best picture of the human soul
(1953/1997, 178). They could object further that his solution to the alleged inability of
outer or inner sense to provide a representation of a thinker/thinking—viz. the
hypothesis of an a priori representation of a cognitive subject as a subject whose states
are necessarily connected—is lazy, extravagant, and has its roots in the metaphysics
he claims to leave behind.
These likely objections can, however, be countered. At one level, Kant agreed with
Wittgenstein’s aphorism. He thought that human form was, in fact, the best evidence
for the presence of a mind.37 As Strawson pointed out long ago, he also recognized
that bodily continuity was the invariable evidence for mental continuity (B415). He
noted in the Anthropology that humans constantly observe the behavior of other
humans and that they should study human behavior to become more astute moral
agents (4.119ff.). The argument for symmetry rests on a narrower claim: In the case of
rational cognition, the predicates applicable through outer sense are irrelevant
(A22/B37, B419-20). They bear no relation to the key notions of ‘rationality’ and
31
‘necessary connection.’ Rational cognizing can be understood only by performing that
activity. For that reason humans can understand other humans to be thinkers only by
using their own activities as models—whence the symmetry of application conditions in
all cases. This argument is compatible with holding that humans can learn a great deal
about what and, indeed, whether others are thinking through observation. What
‘outside’ observation cannot supply is an understanding of what thinking is.
Kant described nativism as a ‘lazy hypothesis’ and tried to differentiate his claims
about a priori concepts from it. Humans are not born with ‘a priori’ representations, but
acquire them through mental activities that are brought into play with the receipt of
sensory data (B1-2). Showing that a representation is ‘a priori’ is not suitable work for
the lazy, since it requires demonstrating that certain mental activities are required for
the production of rational cognition out of the materials of sense. In particular, Kant’s
claim that ‘I-think’ is a priori does not hypothesize an innate representation. What he
takes to be innate in humans is not the representation, ‘I-think,’ but the capacity to form
representations of mental states standing in relations of rational and necessary
connection when the occasion arises (cf. 8.221-23). That capacity is required for
rational cognition. Further, mental activity also produces relations of rational
dependence. When presented with suitable materials, humans have the capacity
consciously to combine some thoughts in others—that they thereby (can) recognize as
rationally and necessarily connected.38 Kant’s claim that humans must be able to
represent mental states as necessarily connected is no additional hypothesis, but a
direct implication of his analysis of rational cognition. If, as I have argued, that analysis
is correct, then it follows that humans must have the capacity to recognize their mental
states as necessarily connected.
The only ‘extra hypothesis’ is Kant’s explanation of the relation between this capacity
and the representation ‘I.’ In addition to logical critiques of various arguments, the
Paralogisms chapter presents an error theory. Traditional metaphysicians assume that
different mental states are necessarily connected to a common ‘I,’ because anyone who
engages in rational cognition must recognize relations of necessary connection across
mental states. Then they hypothesize inherence in an enduring substance to explain
32
the necessary connections. When that extravagant hypothesis is stripped away, we get
back to the real basis of the representation ‘I’: collecting under a single expression the
mental states that stand in the relations of recognized rational dependence that are
necessary for rational cognition. Because those connections are not rooted in
substantiality, but in action, ‘I-think’ is a better expression of the collectivity than ‘I.’ To
say that the representation ‘I-think’ is a priori is to claim nothing more than that humans
must have the capacity to forge and recognize relations of rational dependence across
states and to collect the states so related under a single expression, ‘I-think.’
As we have just seen, Kant’s argument against the asymmetry dogma is narrowly
focused. It does not cover the full range of mental predicates, but only those involving
rational cognition. And rational cognition has a number of special features that may or
may not be generalizable to cases of sensation and desire. Still, it is a sufficiently
important class of cases that it demonstrates that first person third person symmetry
should not simply be assumed when trying to explain self-knowledge and knowledge of
other minds. These cases are also important for my historical thesis. Kant’s theory of
the ‘I-think’ was developed through his analysis of rational cognition in the
Transcendental Deduction and extended in the Paralogisms chapter. What I have just
argued is that those texts do not support the asymmetry thesis, but provide a compelling
argument against it.
Despite his intense study of the Deduction and the Paralogisms, Strawson took
away the wrong lesson about the ‘I-think.’ He thought that Kant’s claims about mental
activity were supposed to provide an explanation of the construction of a known
phenomenal world out of an unknowable noumenal one. Correctly seeing such a
project as incoherent, he dismissed the theory of mental activity as an aspect of the
Critique that had no relevance for current research. By contrast, I’ve argued that the
painstaking study of the mental activity involved in rational cognition has many
interesting implications about how we should understand mental unity, self-knowledge
of belief and how we represent each other.
33
1
Many of these are collected in Shoemaker (1996).
2
Stroud (1968) offered the classic objection; see also his 1994.
3
For a recent example of this assumption, see, e.g., O’Brian, 2007, 127, 158ff.
4
The term comes from Roy Edgley (Moran 2001, 60).
5
Evans’s dispositional approach is supposed to account for how a subject in the sensory deprivation tank
could be correct in referring to himself as an ‘I.’ He is still an ‘I’ just so long as he possesses the right
dispositions (1982, 215-16). On the other hand, it may seem that Evans does not endorse the third
assumption, because in discussing the special features of ‘I’-ideas, he notes that they span time in a
novel way:
An ‘I’-idea gives rise to thoughts dependent upon information received over a period of time …
(which opens up a new possibility of ill-groundedness) (1982, 238).
He develops this aspect of the ‘I-idea’ only in relation to memory, however, and not in relation to
combining information acquired at different times in a present belief. Given this context, he would not
have had Kant’s reason for insisting on self-unity before self-reference (see next section).
6
All references to Kant’s works other than the Critique of Pure Reason are to Kant (1900). References to
the Critique will be given in the text with the usual A/B pagination. The translations are from Pluhar,
except as indicated.
7
I discuss this issue in 2011, Chapter 9, section 3.
8
I use the unwieldy locution, ‘the self-consciousness distinctive to human cognizers’ to mark the fact that
Kant’s argument doesn’t address contemporary questions such as whether animals that pass the ‘mirror’
test are self-conscious. His focus was on rational cognition and the self-consciousness involved in
rational cognition and not on simpler forms that might be shared with animals.
9
I explore the necessity of conscious combining for rational cognition in more detail in section 4.
10
This is not universal. Both Lucy O’Brien (2007) and Christopher Peacocke (2008, Chapter 7) consider
mental act awareness to be pervasive.
11
For example, mathematics and logic seem to be counter-examples to Kant’s claim that all cognition
requires intuitions.
34
12
People sometimes confabulate, offering reasons for judgments that they could not have had, as in the
well-known experiment about choosing among (identical) stockings. Even in this case, however, there is
no suggestion that the subjects were unaware in recognizing all the possible choices as stockings.
Where they confabulate is about the basis of their preferences among them. Confabulation about
preferences does not imply that people are not usually aware in judging, e.g., whether the fruit is ripe. It
may indicate, rather, that they are so accustomed to making (conscious) judgments in light of reasons
that they feel embarrassed at the apparent lapse—and so confabulate. As explained below (note 24),
Kantian judgments are concerned with relatively simple matters, and not with decisions about, e.g., whom
to marry or whether to marry.
13
It is a familiar debate in Kant scholarship whether he argues that the prerequisites of cognition laid out
in the First Critique are needed only for (Newtonian) science or also for ordinary cases of cognition.
14
In a recent paper Hannah Ginsborg (2011) argues that concept use and explicit rule following depends
on a consciousness of ‘primitive’ normativity. In the present case, the idea would be that when very little
children count they have a sense that saying ‘5’ after ‘4’ is what they do; or, with slightly bigger children,
saying ‘5’ and then ‘5’ again is what they do. Ginsborg suggests that, although her account is aimed at
explaining concept and rule acquisition, it also applies to adults (2011, 240). She locates her inspiration
for the idea of ‘primitive normativity’ in Kant’s theory of beauty. Whether it is compatible with his theory of
judgment/concept application (as I have presented it) depends, in part, on whether the feeling of ‘primitive
normativity’ is metacognitive. Given her contrast between humans and parrots, it seems to be. It is less
clear whether the reductionism that she recognizes in her view (2011, 247-249) is consistent with what I
take to be Kant’s insistence that rational cognizers must recognize the necessary connections across
their representations as such (though not in those terms).
15
In his logic lectures and in the Prolegomena, Kant contrasted ‘scientific logic’ with ‘natural logic.’
Ordinary people can sometimes formulate generalizations from their particular uses of rules (natural
logic), but they do not consider, as scientific logic must, the function of rules in making logic or language
possible. See, e.g., 16.14,18, 4.369.
16
Marr’s paradigm of high level theories are the computational theories of e.g. extracting shape from
shading that he and his colleagues worked out in the 70’s and early 80’s. (See my 1988 for further
discussion of this level.) Kant’s analysis of rational cognition has nothing so detailed to offer and could
not be confirmed through computer simulation, which Marr took to be the acid test of computational
theories. Still, it is addressed to the question that computational theories are supposed to answer: What
is the capacity for rational thought a capacity for?
17
In this passage, Kant characterizes the faculty as ‘understanding,’ but he equates the faculties of
understanding and apperception (B133-34n.).
35
18
Although it is possible to combine states with others, it would not be possible continually or ever to
make a grand combination of an entire conscious history. This may be a reason why Kant takes
belonging to a common ‘I’ to require only possible connection to the ‘I-think.’
19
This may seem an odd claim to make since Strawson and many influenced by him have taken Kant to
be a pioneer in addressing the problem that ‘I’ infallibly refers. For reasons I give here and elsewhere
(2011, Chapters 1, 9) I don’t think that is a historically plausible reading of the text.
20
Evans resists the view that that ‘I’ is should be understood in terms of the author of the thought (1982,
213).
21
One might object that it makes no sense to claim that someone could be aware of an action without
being aware of an agent. Although that view is clearly right for the actions of others, it seems that we are
often aware of acting (e.g., very quickly) without any thought of ourselves. We must be implicitly aware of
ourselves, in order to guide the action, but the kind of awareness doesn’t raise issues of circularity.
22
I consider a possible skeptical objection to this account in discussing Evans. See below p. 26.
23
Thanks to Tobias Rosefeldt and Ralf Busse for raising these objections at a workshop on Kant’s
Thinker in Mainz in October of 2010.
24
Given Evans’s parade case of world war three, it seems clear that he is concerned with relatively
straightforward cases of forming beliefs on the basis of evidence and reflection and not with either
unconscious prejudices or deep seated convictions. Kantian judgments are also a separate class from
unconscious representations and matters of faith.
25
I discovered after linking Kant to Lewis Carroll in their understanding of rational cognition that John
Gibbons had done so with respect to Kant’s theory of action (2009, 88). I take a somewhat different
lesson from Carroll than Gibbon does.
26
It might be objected that Frank could just have the thought that wars are brought about by the
preparation for war. Although this is true, if the purported thought is a thought, an example of rational
cognition, then it must be based on some other mental state, e.g. reading Shaw, in which case the same
relation of recognized rational dependence would obtain. One could press further and inquire how it is
possible to have a first thought, but it is far from clear that the notion of a ‘first thought’ or ‘first belief’
makes sense on Kant’s or on anyone else’s understanding of thinking.
27
See my 2011, Chapter 10 (168ff.) and Chapter 14 passim.
28
Here I’m adapting a point that Mathew Boyle (2011) makes in a related context.
29
Kant considers an example very like this in a footnote to the A edition Third Paralogism (A363-64n.)
Since my concerns are systematic and not interpretive, I do not consider the role of that example in his
36
complex account of whether such an individual would be the same ‘substance,’ ‘person,’ or ‘cognitive
subject.’ I treat these issues in 2011, chapter 11.
30
As we will see, it is not clear that Kant would disagree with much of what Evans claims about bodies.
See below p. 31.
31
Evans suggests that the ‘not implausible’ view that self-consciousness requires the conception of
oneself as thinking and as possessing any properties required by thinking can be brought within his
framework of dispositions. To have such a conception is to be disposed to make such judgments as ‘I
was hot and I thought I was hot’ (1982, 259-60). Being disposed to have thoughts may be necessary for
the capacity for rational thought, but it cannot play the role of recognizing relations across thoughts or
recognizing thinking as such that Kant takes to be necessary for rational cognition. See below, pp. 31-32,
for further discussion.
32
Evans’s targets are views like Nagel’s claim that it is impossible to bring the subjective point of view
within the objective world. Although Allison (1996a, 66, 1996b, 103-104) and others have argued that
Kant’s analysis of rational cognition shows that subjects are ‘outside’ of the objective world, I understand
his remarks about the unknowability of the self in a different way. See my (2000 and 2011, 197-200).
33
Ameriks (1982/2000) argues that the Paralogisms should be understood in terms of the metaphysical
views of the Rationialists whom Kant criticized rather than in terms of the cognitive theory of the
transcendental deduction. Still I think the balance of scholarly opinion now assumes that the Paralogisms
chapter needs to be read in light of the theory of the ‘I-think’ in the deduction, and vice versa.
34
Kant believes that all existential propositions require empirical inputs. But he does not infer from this
requirement that the representation ‘I-think’ is empirical, though as the following citation indicates, his
position is not straightforward.
The I think is an empirical proposition, and contains the proposition I exist … when I called the
proposition I think an empirical proposition, I did not mean that the I in this proposition is an
empirical representation. Rather, this representation is purely intellectual, because it belongs to
thought as such. Yet without some empirical representation that provides the material for
thought, the act I think would not take place; and the empirical element is only [nur] the
condition of the application or use of the pure intellectual power (B 422-23n., amended
translation).
I take his point to be that the ‘I-think’ is not an empirical representation, but an a priori one because it is
produced by the activities of the understanding. As we have seen however, such activity can take place
only given sensory inputs, so it follows that the proposition ‘I think’ must also have an empirical basis.
35
I misread the text this way in my 1990.
37
36
Christopher Peacocke offers a general account of the disconnect between understanding and evidence
in cases where subjects’ grasp of a concept has ‘an identity component.’ In the case of rational cognition,
the identity component is using your thinking as a model for thinking. As he notes, such accounts solve
the first person third person ambiguity problem faced by neo-Wittgensteinean accounts. See his 2008,
31-35, 39.
37
In Anthropology from a Pragmatic Point of View, Kant laments that we know rational nature only
through the human species (7.321). His equation of rational creatures and humans is also evident in the
ethics where he refers almost indifferently to humans and to rational beings. This is especially evident as
he lays out the second formulation of the categorical imperative. He glides from talking about how
rational beings must be treated to stating the principle in terms of how ‘humanity’ should be treated
(4.429).
38
Eric Watkins has suggested that this recognition is based on the illusive ‘impression’ of necessary
connection (2005, 274). This seems implausible, because Kant argues that the representation ‘I-think’
must be a priori, just because there can be no sensory impression of necessary connection (e.g. A107).
Bibliography
Allison, Henry (1990). Kant's Theory of Freedom. New York: Cambridge University Press.
_____(1996a) “Kant’s Refutation of Materialism.” In Idealism and Freedom: Essays on Kant's
Theoretical and Practical Philosophy, New York: Cambridge University Press.
_____(1996b). “On Naturalizing Kant’s Transcendental Psychology.” In Idealism and Freedom:
Essays on Kant's Theoretical and Practical Philosophy, New York: Cambridge
University Press.
____(2006). ‘Kant on Freedom of the Will.’ Cambridge Companion to Kant and Modern
Philosophy., P. Guyer, ed., Cambridge University Press: 381-415.
Ameriks, Karl (2000). Kant's Theory of Mind. Second Edition, Oxford: Clarenden Press.
Anscombe, G.E.M.A. (1958/1979). Intention. Ithaca, N.Y.: Cornell University Press.
________. (1975). “The First Person.” In Samuel Guttenplan, ed., Mind and Language (Oxford:
Clarendon Press, 1975), pp. 45-65.
Arnauld, Antoine (1662/1964). The Art of Thinking. J. Dickoff and P. James, trans., New York:
Library of Liberal Arts.
Boyle, Matthew (2011). “Transparent Self-Knowledge.” Proceedings of the Aristotelian
Society, Supplementary Volume 85: 223-41.
Burge, Tyler (1996). “Our Entitlement to Self-Knowledge.” Proceedings of the Aristotelian
Society 96:91-116.
38
_____ (2010). Origins of Objectivity. Oxford University Press.
Carroll, Lewis (1895). “What the Tortoise Said to Achilles.” Mind, 4: 278-80.
Evans, Gareth (1982). The Varieties of Reference, J. McDowell, ed., Oxford: Clarendon Press.
Gibbon, John (2009). “Reason in Action.” In Lucy O’Brien and |Matthew Soteriou, eds., Mental
Actions. Oxford University Press: 72-94.
Ginsborg, Hannah (2011). “Primitive Normativity.” Journal of Philosophy, CVIII, no.5: 227-54.
Kant, Immanuel (1781.1787/1966). Critique of Pure Reason: Unified Edition. Pluhar, Werner,
trans. Indianapolis: Hackett.
______ (1900- ). Kants gesammelte Schriften, Akademie Ausgabe, edited by the Koeniglichen
Preussischen Akademie der Wissenschaften, 29 vols. Walter de Gruyter and
predecessors, 1902-.
Karmiloff-Smith, Annette (1992) Beyond Modularity: A Developmental Perspective on
Cognitive Science. Cambridge, Mass.: MIT Press/Bradford Books.
Kitcher, Patricia (1988). “Marr’s Computational Theory of Vision.” Philosophy of Science,
55:1-24.
_______ (1990). Kant’s Transcendental Psychology. New York: Oxford University Press.
______ (2000). “On Interpreting Kant’s Thinker as Wiggenstein’s ‘I’.” Philosophy and
Phenomenological Research LXI, no. 1: 33-63.
______ (2011). Kant’s Thinker. Oxford University Press.
Locke, John (1690/1975). Essay Concerning Human Understanding. P. H. Niddictch, ed.
Oxford: Clarendon Press.
Marr, David (1982). Vision. San Francisco: W. H. Freeman.
Mele, Alfred (2009). “Mental Action: A Case Study.” In Lucy O’Brien and |Matthew Soteriou,
eds., Mental Actions. Oxford University Press: 17-37.
O’Brien, Lucy (2007). Self-Knowing Agents. Oxford University Press.
Peacocke, Christopher (2008). Truly Understood. New York: Oxford.
Rӧdl, Sebastian (2007). Self-Consciousness, Cambridge, MA: Harvard University Press.
Shoemaker, Sydney (1996). The First Person and Other Essays. Cambridge University Press.
Strawson, P. F. (1966). The Bounds of Sense. London: Methuen.
Stroud, Barry (1968). ‘Transcendental Arguments.’ Journal of Philosophy: 241-56.
39
___ (1994), ‘Kantian Arguments, Conceptual Capacities, and Invulnerability. ’ Kant and
Contemporary Epistemology. P. Parrini, ed., Kluwer.
Watkins, Eric (2005). Kant and the Metaphysics of Causality. New York: Cambridge University
Press.
Wittgenstein, Ludwig (1953/1997). Philosophical Investigations. G.E.M. Anscombe, trans.,
Oxford: Basil Blackwell.
40