You are on page 1of 31

Synthesis, Cognitive Normativity, and the

Meaning of Kants Question,


How are synthetic cognitions a priori
possible?
R. Lanier Anderson

Kant organizes the critical philosophy around the question, How are synthetic cognitions a priori possible? (Prol. 278),1 which he took to have enormous importance:2
The real problem of pure reason is now contained in the question: How
are synthetic judgments a priori possible?
That metaphysics has until now remained in such a vacillating state of
uncertainty and contradictions is to be ascribed solely to the cause that no
one has previously thought of this problem. . . . On the solution of this
problem. . . metaphysics now stands or falls. (B 19)
Attention to Kants question usually focuses on the specific nature of the synthetic
a priori, but in my view there are also two broader issues in play. First, Kant
believes there is a special problem about synthetic judgment as opposed to
analytic judgment, which turns on a simple logical relation of concepts (viz.,
containment). Second, there is a problem about the possibility of knowledge as
such. The present paper explores this last aspect of Kants question. I argue that
it is meant to raise fundamental issues about the normativity of cognition, in a way
that sheds light on the most central and characteristic arguments of the Critique.
Kant asks his question for the benefit of metaphysics, which purports to
contain synthetic a priori cognition. The hope is to learn something about such
cognition in general by investigating the uncontested (Prol., 275) cases of mathematics and exact natural science. The general features of actual scientific cognition are then supposed to indicate what genuine metaphysical knowledge would
have to look like, in part by revealing what gives cognition its normative force
i.e., what makes it valid and binding for belief. With such an account in hand, we
can finally set metaphysics on the secure course of a science (B vii), by gaining a
deeper understanding of how normative knowledge is possible at all.
Following a widespread tendency in early modern philosophy, Kant views
cognition as a product of the powers and activities of the mind (see section 4,
below). The approach appears obvious enough at first glance, but from our
current point of view, it seems to render cognitive normativity harder to understand, for reasons connected to the now-familiar charge of psychologism. This is
especially true once we restrict attention to synthetic claims, which cannot inherit
European Journal of Philosophy 9:3 ISSN 09668373 pp. 275305 Blackwell Publishers Ltd. 2001. 108 Cowley Road, Oxford
OX4 1JF, UK, and 350 Main Street, Malden, MA 02148, USA.

276

R. Lanier Anderson

normative force from purely logical relations among their concepts. The story I
will present suggests that in fact, later worries about psychologism were decisively shaped by Kants question about cognition. It made visible basic assumptions about normativity built into the early modern conception of mind, whose
previous defense was not satisfactory. Kants own agenda was not to undermine
these assumptions, however, but to explain knowledge within the early modern
framework, by appealing to actions of the mind. As a result, he had to address
two main points. First, he needed an account of the nature of the mental operations that generate normative cognition. The Critiques detailed theory of synthesis, discussed in sections 2 and 3, fulfills this function, and simultaneously
provides the model in terms of which Kant executes the works main arguments.
Second, if the resulting knowledge is to count as true, Kant must explain how
cognitive synthesis is reliably connected to the objects of knowledge. I will argue
in section 5 that the idealist Copernican revolution supplies his response to this
challenge, thereby completing the explanation of cognitive normativity begun by
the theory of synthesis.
It is thus no accident that Kant thought his organizing question was so important. It leads directly into the core arguments of the Transcendental Analytic
the arguments in which Kant defends his systematic account of nature and our
knowledge of it. My aim here is to uncover a basic argumentative strategy that
guides Kants procedure throughout the Analytic. This pattern of reasoning ties
together three of Kants principal doctrines, all of which are often dismissed
nowadays as indefensible: 1) the view that mental actions of synthesizing representations are central to the explanation of knowledge, 2) Kants account of mathematical proof, and 3) transcendental idealism. The resulting interpretation
makes a certain amount of sense of these seemingly embarrassing views, and
shows why Kant thought they were so central. Understanding the importance he
attributed to these ideas will also clarify some of our own philosophical problems, and indicate the theoretical costs of some elements of our current philosophical common sense.

1. Two Alternative Readings of Kants Question


We can frame our problem by reference to two broad approaches to understanding Kants question, which were well developed in the nineteenth century, and
also have twentieth century adherents. One offers a psychological reading of
Kants Critique, the other an epistemological reading. Each approach has something
to recommend it, but both also have difficulties. I will seek to build on their
strengths in my interpretation below.
The psychological reading understands Kant to be offering a theory of the
workings of the cognitive mind. This idea (recently revived by Patricia Kitcher
1990 and others3) provides a natural way to understand Kants how possible
question viz., it is a question about the ways and means of cognition. To ask
how knowledge is possible in this sense is to ask what the mechanisms are by
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

277

which the mind produces knowledge. Since the nineteenth century, readers in
this tradition have often criticized Kant on the psychological details, but they
praised him for opening up a new method for philosophy, founded on discovering the features of experience that are due to our own physio-psychological organization.4 Many of them have also sought to correct and update the critical
philosophy based on new discoveries. For example, F. A. Lange treats
Helmholtzs doctrine of unconscious inference as empirical confirmation of
Kants hint that sense and understanding might have some common but to us
unknown root (A 15/B 29), as well as of the basic Kantian idea that logical structure is imposed onto experience by the setup of the mind (Lange 1876, II, 312).5
There is substantial textual support for such an approach, taken broadly. In the
key arguments of the Analytic, Kant clearly means to be offering a theory that
explains how the mind manipulates representations (e.g., concepts, intuitions) to
produce cognitions. Kants central arguments appeal to mental actions of synthesis, or combination, which on his view forge real connections among our representations (connections of the sort whose possibility was doubted by Hume in his
famous dictum that all our ideas are loose, or unconnected).6 The ubiquity of
this explanatory strategy in Kant is widely acknowledged, even by philosophers
who find Kants indulgence in psychology a source of embarrassment.7
One thought that generates resistance to the psychological reading is that
Kants strategy purports to reach a normative result it attempts to underwrite the
justification of valid knowledge like geometry or natural science. These normative
intentions raise difficulties for a psychological interpretation of Kants theory of
cognition, which would explain knowledge through mental processes, like
combining the parts of a perception, or forming a belief through Humean laws of
association. Since Kant treats cognitions as individual mental states, it is natural
to assume that the posited psychological processes explain cognitions by causing
them an assumption that fits well with Kants understanding of empirical
psychology as a causal theory based ultimately on simple laws of association. But
precisely that fact should give us pause, for Kant himself insists that his theory of
cognition does not belong within empirical psychology, so conceived.8
Indeed, Kants anti-psychologism is based on the thought that merely psychological explanations could never account for the normative standing of cognitions.9 The difficulty arises from a fundamental structural difference between
naturalistic and normative rules. Unlike a descriptive natural law, a prescriptive
normative rule does not entail that all the particular cases it covers actually
conform to the rule. If some cases violate a purported natural law, we conclude
that the law was mistaken, and we adjust it to fit the new facts. By contrast, when
an event violates some normative rule, we nevertheless hold it accountable to the
rule, and count it as wrong or blameworthy because it does not conform. The
normative rule thus remains binding, even when it is violated, and thereby has a
different direction of fit from descriptive rules.
Psychological processes are described by naturalistic causal laws, not prescriptive normative rules; if some causal account predicted the emergence of a particular cognitive state, but a different one occurred instead, the right theoretical
Blackwell Publishers Ltd. 2001

278

R. Lanier Anderson

response would be to count the causal hypothesis as disconfirmed by the experiment, not to blame the actual cognition for being wrong.10 Thus, the causal explanation of a cognition does not account for its normative force. To do that, we need
an explanation compatible with various outcomes, which can retain its validity
even if a false cognition is produced, and thereby underwrite our judgment that
the actual cognition is wrong. This basic idea was developed and deployed in the
late nineteenth century as the charge of psychologistic fallacy the fallacy of identifying a normative rule of reasoning or cognizing with an exceptionless, descriptive natural law.
In recent decades, it has been fashionable to make the psychologism charge
against early modern philosophers, including Kant (see, e.g., R. Rorty 1979:
14855, esp. 1512). But in fact, as neo-Kantians emphasized already in the nineteenth century, this was Kants own point against empiricism.11 The empiricists
offered a theory of the natural laws of cognition a physiology of the understanding (A ix; cf. A 857/B 1179) and Kant shows that, whatever they discovered about the way our concepts emerge in fact, this would still leave the further
question with what right we can use them to produce justified cognitions. That is,
empiricist psychology fails to account for the normativity of cognition, and the
outstanding normative question of right is just the one Kants transcendental
deductions were meant to address.12
So here is the problem for a psychological reading of Kant. Given the model of
science which guides the eighteenth century debate, empirical psychological
processes must be described by exceptionless laws, ultimately unified under a
few, simple principles. Kant was as aware as anyone that, given this conception
of natural law, there is a deep, in principle difference between questions belonging to the normative and to the naturalistic realms: for him, the ought, if one has
merely the course of nature before ones eyes, has no significance whatever (A
547/B 575). Therefore, psychological readers must be wrong that Kant intended
to explain knowledge by appeal to empirical psychological processes.
Such considerations led to a purely epistemological reading of Kants question
about how knowledge is possible.13 On this reading, the question is not about
how the mind generates representations. Instead, it concerns objective relations of
justification between, on the one hand, bodies of knowledge conceived of as
mind-independent abstract objects, and on the other hand, the world they are
true of, or the data that are evidence for them. Within the epistemological interpretive tradition, Kants regressive method has assumed great importance. The
method generates transcendental arguments, which start from a given body of
knowledge (or practice), which is assumed to be valid, and then infer that there
must be some transcendental structure that explains the normative force of the
assumed body of knowledge (or practice) in Kant himself, this would be a transcendental faculty of the mind. So, e.g., Kant starts from our valid knowledge of
geometry, and explains that knowledge by claiming that space is the form of
outer intuition.
Despite the mentalistic overtones of Kants own conclusion, the starting
point of the argument (the initial body of knowledge) can be characterized as a
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

279

mind-independent achievement of the culture the validity of geometry does not


depend on its being apprehended by any particular person.14 It was therefore
possible for anti-psychologistic readers of Kant to sever the mentalistic elements
from his view; all they had to do was to refuse to posit faculties of the transcendental mind to explain normativity, as Kant had done. Various later Kantians
have offered many different proposals for an anti-psychologistic basis for normativity, once Kants appeals to the transcendental mind have been abandoned (e.g.,
the transcendental values of the neo-Kantians, fundamental epistemic conditions,
etc.). Often these proposals raise philosophical problems of their own, but the
important point for our purposes is just that positing a non-mental version of the
transcendental allows Kants theory of mental actions to be submerged. Such
anti-psychologistic readings have dominated Kant commentary since the late
nineteenth century.15
The limitation of the epistemological approach, however, is just the flip side of
the strength of the psychological reading. When the epistemological reading
disavows the theory of mental activity Kant offers in the Critique, it is forced to
read away, or to criticize as misguided, an enormous amount of the account he is
offering.16 The danger here is that Kants actual arguments will be suppressed
along with the theory of mental actions.
To conclude, both the psychological and the epistemological readings get
something right about the Critique. The psychological reading is right that Kant
offers us a theory of the mental actions involved in cognition, and the epistemological reading is right that he is trying to provide a normative account of cognitions. But that is just the puzzle. Precisely the epistemological readers emphasis
on the normative status of Kants account seems to rule out the psychological
readers insistence on the theory of mental action. Does this mean that Strawson
was right to complain that the core doctrine of the Transcendental Analytic at
least in Kants own hands is incoherent in itself (Strawson 1966: 16), because of
the way it appeals to mental activities of ordering experience? The Strawsonian
reaction is too quick. Before despairing of the mentalistic arguments of the
Transcendental Analytic and resorting to a radical reconstruction that casts
Kants ideas in substantially different form, we should try to understand his
appeals to mental processing on their own terms. In the next three sections, I will
outline an approach to the Analytic that tries to accommodate the advantages of
both epistemological and psychological readings. While the resulting alternative
account raises deep philosophical issues of its own, I will suggest that these are
just the challenges Kant meant to be pressing on us.

2. Kants Doctrine of Synthesis in the Argument of the Analytic


The core idea in Kants talk of cognitive mental processing is the notion of synthesis. By investigating its argumentative role, we can more carefully evaluate the
possibility that claims about mental processes can legitimately have implications
for a normative theory of cognition. Kant officially introduces synthesis in the
Blackwell Publishers Ltd. 2001

280

R. Lanier Anderson

first chapter of the Analytic, on the Clue to the Discovery of all Pure Concepts of
the Understanding (A 6683/B 91116): synthesis in the most general sense . . .[is]
the action of putting different representations together with each other and
comprehending their manifoldness in one cognition (A 77/B 103). According to
this definition, the full-scale cognitive process of synthesis involves three
elements: 1) a manifold of content to be combined (the different representations);
2) an action of combining the manifold (the putting together); and 3) a representation of unity (one cognition; emph. added), which serves as the principle of
order connecting the manifold content (see A 789/B 104; also B 1301).17 So
synthesis combines a given manifold into a unified whole.
Such combinations are supposed to be required for us to produce genuine
cognitions out of representations with manifold content. The thought is that any
representation rising to the level of cognition will be complex, and will exploit that
complexity in making its cognitive claim. Therefore, cognitions depend on a
mental capacity to represent the various aspects of some complex content separately (and thus explicitly), and then to link them to one another in a way that
(likewise explicitly) represents their relations, making all the content available for
cognitive duty.
We can now ask what kind of cognitive power synthesis is, or, in Kants terms,
to what faculty the activity of synthesis belongs. The standard answer is Kants
oft-quoted dictum that synthesis is the mere effect of the imagination, of a blind
though indispensable function of the soul . . . of which we are seldom even
conscious (A 78/B 103). The blindness highlighted here suggests that synthesis
is simply a psychological tendency to associate representations according to
(Humean) laws. But if full cognitive synthesis really involves all three of the
elements mentioned above, then it must be blind only in some restricted sense.
And indeed, the broader context of the blindness passage itself suggests a richer
account of the full-scale synthesis that generates cognitions. Kant writes,
The synthesis of a manifold, . . . first brings forth a cognition . . .; the
synthesis alone is that which properly collects the elements for cognitions
and unifies them . . . . Synthesis . . . [is] the mere effect of the imagination,
of a blind though indispensable function of the soul, without which we
would have no cognition at all, but of which we are seldom even
conscious. Yet to bring this synthesis to concepts is a function that pertains
to the understanding, and by means of which it [the understanding] first
provides cognition in the proper sense. (A 778/B 103)
Kant begins from the idea that synthesis produces cognitions, because it alone
can unify representations in the right way. He then classifies synthesis as an effect
of blind imagination, and counts imaginative processing as a necessary condition
of cognition. But Kant immediately introduces the qualification that cognition in
the proper sense requires that synthesis be guided by concepts provided by the
understanding. So it is the understanding, and not the blind imagination by itself,
that is sufficient to first provide genuine cognition (which was the achievement
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

281

attributed to synthesis alone in the initial sentence). It follows that synthesis


cannot be a mere effect of blind imagination after all, if it is to do the work of
first bring[ing] forth a cognition, and indeed do it alone. The appearance of
inconsistency in Kants usage can be removed by considering the remarks he
offers further down the page:
The first thing that must be given to us a priori for the cognition of all
objects is the manifold of pure intuition; the synthesis of the manifold by
means of the imagination is the second thing, but it still does not yield
cognition. The concepts that give this pure synthesis unity, and that
consist solely in the representation of this necessary synthetic unity, are
the third thing necessary for cognition,. . . and they depend on the understanding. (A 789/B 104; cf. B 1301)
Here Kant explicitly teases apart the three elements that go into any complete
synthesis. Synthesis by means of the imagination is only one of those elements,
and since Kant insists that it still does not yield cognition, it must be only a
restricted aspect of the full action of the mind that first brings about a cognition.
We should therefore see Kant as implicitly distinguishing between, on the one
hand, synthesis as a mere effect of imagination, and on the other, synthesis as an
action of the understanding (B 130),18 sufficient to unify constituent representations into a cognition. The former is (only) one element of the latter, because full
synthesis combines synthesis in the former, thinner sense with the idea of the
manifold synthesized and (crucially) a conceptual representation of the unity of
the synthesis.
These are not merely arcane points of Kants faculty psychology. As Kitcher
notes, Kant relies extremely heavily on the notion of synthesis to state his theory
of cognition: she counts over sixty references to it in the A Deduction alone
(Kitcher 1990: 74). The philosophical question here is whether synthesis is fundamentally conceptual (i.e., dependent on the rule-following understanding), or
merely causal (an act of blind, rule-described imagination). Synthesis in the
complete, three part sense essentially involves a conceptual representation of
unity, which serves as a rule for the minds combining activity (A 789/B 1034).
It can therefore be understood as a kind of function that takes other representations as inputs, and yields as output a new representation that combines contents
derived from the inputs according to a characteristic pattern (fixed by the
concept).19 It thereby creates not merely a co-locating of different representations
in the mind, but a genuine melding of their contents. This is quite different from
the mere association of representations by Humean laws, which does not reorder
the contents within associated representations in any way.20 The role of conceptual rules for unity is suggestive for our purposes; it may be that precisely here
space has been opened for synthesis to do normative work (e.g., if the rules could
receive some normative interpretation). I will return to this suggestion below.
The concept-laden conception of synthesis lends Kants theory of cognition a
top-down flavor. Some commentators therefore resist it, in order to read Kants
Blackwell Publishers Ltd. 2001

282

R. Lanier Anderson

account as a bottom-up explanation of how the mind builds higher cognitions


on the basis of more rudimentary cognitive processes.21 The cited texts make it
clear, however, that the activity of synthesis does essentially involve higher
concepts. In addition, Kants most characteristic arguments have a fundamentally
top-down structure, aiming to show that even rudimentary cognitive operations
like perception are actually parasitic on the full conceptual activity he attributes to
the understanding.
One such argument appears in the immediate context of Kants introduction of
synthesis. Just after the quoted passages, Kant deploys the notion of synthesis in
a pregnant sketch that outlines his basic strategy in the Transcendental Analytic.
The aim of Kants Clue chapter is to exploit the logical functions of judgment as
the guiding thread [Leitfaden], to the discovery of a system of the fundamental
concepts of the understanding, or categories.22 The thought is that the understanding characteristically works by forging connections among representations
in judgments, and that the system of logical forms of judgment will therefore give
rise to categories that the understanding uses to guide synthesis. If Kant could
show that these categories are required for all synthesis, and that all experience
(even down to perception) rests on such syntheses, then it would follow that the
categories are conditions of the possibility of experience.
The details of this argument are notoriously devilish, and Kant reaches them
only in the ensuing Transcendental Deduction. (Indeed, the application of the
strategy to particular categories awaits the still later arguments of the Analytic of
Principles.) Nevertheless, Kant sketches his general idea already in the Clue
chapter, by way of motivating the importance of his table of categories. Here is
what he writes:
The same function that gives unity to the different representations in a
judgment also gives unity to the mere synthesis of different representations in an intuition, [and it] . . . is called the pure concept of the understanding. The same understanding, therefore, and indeed by means of
the very same actions through which it brings the logical form of a judgment into concepts. . . also brings a transcendental content into its representations by means of the synthetic unity of the manifold in intuition. . .,
on account of which they are called pure concepts of the understanding
that pertain to objects a priori. (A 79/B 1045)
This text reiterates the thought that the categories are to be understood as functions of synthesis that give unity to different representations. But the key idea is
that the same synthesis provides both the unity of a full conceptual judgment, and
the unity of representations in an intuition, or perception. That is, the lower level
synthesis involved in perception is already at the same time a higher level cognitive synthesis (guided by concepts of the understanding, and potentially warranting a full-scale cognitive judgment). Thus, all experience presupposes the
synthesizing role of the understandings categories, and so they cannot be
derived from experience.
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

283

Kant does not say merely that synthetic processes of the same general sort play a
role in both perception and higher level conceptualization, but that the synthesis
is numerically identical. The two tasks are carried out by the same instance of
synthesizing: the same understanding by means of the very same actions (A 79/B
105; emph. added) brings about both kinds of unity. Indeed, it is because the same
synthesis can do all this, that we can say the categories pertain to objects a priori
(A 79/B 105). In my view, this is the fundamental idea behind the arguments of
the Analytic, and I will call arguments that turn on it same synthesis arguments.
There are questions to raise here. The synthesis of concepts in a judgment and
the synthesis of intuitive elements into one perception appear to be radically
different cognitive tasks, calling for wholly different kinds of synthesis. What can
Kant mean by asserting that these tasks are brought off by the same act? It is also
far from evident how this identity suffices to show that the categories pertain to
objects a priori. Perhaps most troubling, Kant himself differentiates the kinds of
synthesis from one another, apparently contradicting what I have just called his
fundamental idea. The most famous example is probably the discussion of the
threefold synthesis (A 97) in the A Deduction, where Kant distinguishes a
synthesis of apprehension in intuition (A 98100), from a synthesis of reproduction in imagination (A 1002), and a synthesis of recognition in a concept
(A 10310).
The last contradiction is merely apparent, however. In the A Deduction, Kant
speaks of a single threefold synthesis, not three different acts or kinds of synthesis. His argument there repeatedly returns to the idea that the different levels of
synthesis are inseparably combined (A 102), because they are ultimately identical in some way (see A 108, 118, 119). I will argue that Kant distinguishes different levels in a single synthesis, which should be understood as more and less
abstract structures all contained in and realized by the single given act. More
abstract structural features provide the form of the synthesis, which is then filled
in by the more concrete details. We can clarify the relation between these different levels, and also see how Kants claim about the identity of the synthesis is
supposed to help establish the a priori objective validity of the categories, by
investigating the role of synthesis in one of Kants arguments.

3. How Same Synthesis Arguments Work, and the Nature of


A priori Synthesis
Consider the Axioms of Intuition, where Kant argues that the categories of
quantity (and therefore also the results of mathematics) are applicable to all
appearances, because they are preconditions of perception. Kant begins by pointing out that the perception of any extended thing, e.g., an elliptical table top, is a
representation with complex content. Perception represents the table top as
having various parts, for example. So there must be a synthesis that goes through
and combines the complex content, in order to produce a single cognition that
explicitly represents the several parts in their interrelations. Next, Kant observes
Blackwell Publishers Ltd. 2001

284

R. Lanier Anderson

that, since the perception represents the parts of the table top as located in space
(itself a complex content), the perceptual synthesis depends on an underlying
synthesis of the space the table is in. That synthesis combines a manifold without
regard to qualitative differences of content (i.e., it is a manifold of the homogeneous (B 203)). Finally, as we saw, Kant insists that every full cognitive synthesis
involves a concept, i.e., a function which abstractly represents the whole into
which the synthesized elements are combined, and provides a rule guiding the
combination. The most general concept for our case is a rule combining homogeneous units, viz., the concept of quantity, or magnitude. Thus, Kant claims, the
synthesis of space itself depends on a still more abstract synthesis according to
this concept. In short, then, Kants argument is that 1) the perception of the table
top is a synthesis, which 2) depends on the underlying synthesis of space, which
in turn 3) requires a synthesis according to the concept of quantity. Therefore,
applying the categories of quantity to experience is a precondition even of
perception, and so all appearances are quantities.
The crucial move in the argument is its assertion of a top-down dependence
of perception on the synthesis of space, and ultimately on the synthesis according
to the categories of quantity. The same synthesis strategy provides the rationale
for this direction of dependence. As I reconstructed the argument, Kant considers
three distinct levels of synthesis: 1) the full perceptual synthesis of the table top,
2) the synthesis of the space the table top fills (here an ellipse), and 3) the synthesis under the concept of quantity. According to the same synthesis idea, however,
these are not three syntheses, but only one synthesis with three levels, related as
form to matter, and that is why the lower (perceptual level) synthesis depends on
a higher conceptual synthesis. Here is Kants own version of the conclusion:
Thus even the perception of an object, as appearance, is possible only through
the same synthetic unity of the manifold of given sensible intuition through
which the unity of the composition of the homogeneous manifold is
thought in the concept of magnitude, i.e., the appearances are all magnitudes. . . because as intuitions in space and time they must be represented
through the same synthesis as that through which space and time are
determined. (B 203, ital. added)
When we perceive the table top, then, we simultaneously make a synthesis of an
elliptical space, which is nothing but an abstract structural feature of the very
empirical synthesis involved in perception. (The table top (qua appearance) is
represented through the same synthesis . . . through which space . . . [is] determined.) Likewise, the synthesis according to the category of quantity is an
abstract structure of that same empirical synthesis (the same synthetic unity).
The more abstract structures of the synthesis (the form) are filled in by more
concrete details (the matter) to make the full empirical synthesis. Qua forms, the
abstract structures constrain the shape taken by the empirical content, while the
empirical content provides a specification of the abstract structures. Thus, lower
syntheses depend on higher ones, which determine their form.
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

285

Since the synthesis that produces cognition of this table is the same synthesis
that produces cognition of an ellipse (only now filled in with sensory content) the
mathematical truths we can derive about the ellipse will be literally true of the
table, and indeed, a priori true of it. For instance, the area of the table top must be
times the product of the radii, because that result characterizes all elliptical
spaces a priori, and the synthesis of apprehension involved in the perception of
the table top is just a synthesis of an elliptical space, filled in with some particular
content, or matter of sense (A 20/B 34). The categorial result that the table has a
quantity is a priori true for parallel reasons. Kants argument, then, is that one and
the same action of synthesis produces both the apprehension of an object, and at the
same time the more abstract achievements of generating the spatial framework,
and subsuming the appearance under the category of magnitude. Because the
fully concrete apprehension is just the filling in of these more abstract structures,
it follows that they apply (with objective validity) to the apprehended object.23
If I am right about Kants argumentative strategy, then his notorious a priori
synthesis cannot be a mysterious, sui generis action of the mind, separate from
ordinary empirical synthesis, as is often assumed.24 Rather, it is a structural
feature of the very empirical synthesis that produces experience. The a priori
structure of synthetic activity is isolated from the full empirical synthesis by
abstracting from the details of the sensory matter being synthesized an abstraction legitimated by the basic form/matter distinction that organizes Kants theory
of cognition.25 In the Anticipations of Perception, Kant describes it this way:
Perception is empirical consciousness, i.e., one in which there is at the
same time sensation. . . . Appearances therefore. . . contain. . . the materials for some object in general. . ., i.e., the real of sensation. . . . Now from
the empirical consciousness to the pure consciousness a gradual alteration is possible, where the real in the former entirely disappears, and a
merely formal (a priori) consciousness of the manifold in space and time
remains. . . . (B 2078)
By such a gradual abstraction, according to Kant, we eventually arrive at the
formal, or structural, features of synthesis which are not derived from or dictated
by the matter given in sensation; that is, they are independent of sensation, or a
priori. Once the theory of cognition has uncovered these a priori features via
abstraction from paradigm cognitive syntheses, then we can also go in the reverse
direction to reconstruct experience. Reconstruction starts from the abstract structure provided by metaphysical categories and mathematical truths, and fills in
that schema with sensory matter to reach full blooded experience.
So far this talk of filling in a schematic structure remains somewhat
metaphorical, and some readers may resist its use of abstraction as too empiricist.
We can render it more specific by appeal to ideas from Kants philosophy of
mathematics. While it may not seem initially obvious, a detour through the
philosophy of mathematics makes perfect sense under Kants account of the organizing question of his philosophy. The point was to ask how synthetic cognition
Blackwell Publishers Ltd. 2001

286

R. Lanier Anderson

a priori was possible in the actual cases (mathematics, exact natural science), so as
to use what was learned from those examples to improve the prospects for metaphysics. Extending cognitive strategies proper to mathematics into a new use in
metaphysics is thus a direct application of the general approach.
The relevant mathematical notion is construction.26 Kant adduces mathematical
constructions as his leading examples of schemata for synthesis in his chapter on
the schematism (A 1401/B 17980), and in my view, he understood all a priori
synthesis by analogy to mathematical constructions like geometrical diagrams.
Their function in proof offers a clear paradigm for what it is to fill in an abstract
structure to produce a more concrete representation. For example, in Bk. I, Prop.
1 Euclid describes the procedure for constructing an equilateral triangle, which
exploits two intersecting circles constructed on the same line segment, each with
the line segment as radius and one end point as center. Because the construction
procedure abstracts away from particular determinations like the magnitude of
the initial line segment, it provides a schematic representation of an equilateral
triangle, which applies equally well to all such triangles. Particular equilaterals
can be understood as instantiations of the construction procedure, which translate its schema into an image (A 140/B 17980) by rendering concrete the various
details (e.g., side size) from which the a priori construction rule abstracts.27
It might be objected that mathematical construction is a completely a priori
matter, whereas we were concerned to illuminate the relation between a priori
and empirical levels of synthesis. But Kants account of how constructions can
produce a priori and necessary mathematical results clarifies just this relation.
Consider the following passage:
The construction of a concept . . . expresses . . . universal validity for all
possible intuitions that belong under the same concept. Thus, I construct
a triangle by exhibiting an object corresponding to this concept, either
through mere imagination, in pure intuition, or on paper, in empirical intuition, but in both cases completely a priori, without having had to borrow the
pattern for it from any experience. The individual drawn figure is empirical, and nevertheless serves to express the concept without damage to its
universality, for in the case of this empirical intuition, we have taken
account only of the action of constructing the concept, to which many determinations, e.g., those of the magnitude of the sides and angles, are
entirely indifferent, and thus we have abstracted from these differences. . .
(A 71314/B 7412, ital. added)
Clearly, the construction of an empirical figure with pencil and paper is an empirical process an empirical synthesis. Nonetheless, the construction manages to
express a universal, a priori result, because we abstract from the empirical details
of the synthesis and attend only to the form of the action of the mind, by which it
constructs the concept. This schema (see A 1402/B 17981) acts as a rule of
synthesis. It governs how the construction must be carried out, and is valid a
priori, and therefore universally for all the concrete empirical instantiations. That
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

287

is, the construction rule is a schematic form which can be filled in by material in
different particular ways, while retaining its identifying structural features.28 Such
construction rules stand to the empirical, pencil and paper constructions, just as
pure a priori synthesis stands to empirical synthesis.
Indeed, Kants reference to the mental action of constructing shows that
construction rules just are abstract a priori aspects of the syntheses involved in
actual empirical constructions. Moreover, they are automatically valid of all the
concrete empirical syntheses that generate diagrams instantiating them, precisely
because the empirical diagrams are just fillings in of the relevant schemata. The
same considerations account for Kants apparently absurd claim that empirical,
pencil and paper constructions are a priori in some sense: they are so because they
have an abstract structural aspect (captured by the construction rule) that is a
priori, and the only important thing about the empirical action of construction is
its status as an instance of that a priori structure of synthesis. (We abstract from
all the rest.) Mathematical construction thereby exemplifies the form/content
relation between the pure and empirical syntheses which Kants same synthesis
arguments exploit to prove the validity of metaphysical concepts.
The notion of a schema thus turns out to be of crucial importance for understanding Kants procedure in the Transcendental Analytic.29 Kant opens the
Analytics final chapter claiming to have established that The principles of pure
understanding . . . contain nothing but only the pure schema, as it were, for possible experience (A 2367/B 296). The model is worth taking seriously, for the categories provide the form for empirical synthesis (and thus to experience) in much
the way a schematic construction dictates the form of a geometrical object. The
categories are of course more abstract than any geometrical construction so
abstract, indeed, that they require a special investigation to produce the appropriate schemata for relating them to spatio-temporal experience.30 But once the
schematized category is in place, it functions in a directly analogous way. Kant
says it provides a monogram (A 142/B 181) (in the sense of a simple outline
drawing), which serves as a rule of the synthesis (A 141/B 180). That synthesis,
in turn, fills in the monogram by producing a more specific, concrete instantiation
of the schematic form it represents.
It is time to sum up the result of this section. Kant understood the categories as
rules for a priori synthesis. The a priori part of synthesis provides the form of our
experience, and constrains the shape taken on by the empirical parts. That is, the
most abstract concepts of metaphysics, and below them the a priori structures of
mathematics and exact mathematical natural science, provide a schema that is
gradually filled in by more concrete and articulated structures of synthesis as we
descend through it. This a priori schema is eventually itself filled in with additional detail by the sensory matter of experience, insofar as that matter is
subsumed and organized by the concepts of the special sciences. In this sense, I
think Michael Friedman is exactly right to suggest that the regulative ideal
toward which Kants metaphysics of nature aims is a wholly systematic, but still
abstract, picture of the natural world, presented as a kind of construction (analogous to a mathematical construction), to be filled in by experience.31
Blackwell Publishers Ltd. 2001

288

R. Lanier Anderson

Thus, the critical philosophy of nature might be figured by a map printed on a


series of transparency pages, where the coastal/border outline is printed on the
bottom page, and each additional page carries features of some particular type
(elevation, river systems, vegetation cover, roads and towns, population, etc.).
The additional features are added to the total picture as we fold the transparency
pages down over the basic outline map at the bottom. The outline dictates the
general shape (and thereby constrains the range of particular shapes) that may be
taken by more specific features. But the abstract, formal parts of Kants map the
a priori parts of synthesis have objective reality only because they are just high
level structures of the very same empirical syntheses through which actual objects
are cognized. As Kant puts it, empirical synthesis is . . . the only kind of cognition that gives all other synthesis reality (A 157/B 196). That is, a priori categories
and mathematical truths are valid of objects because they are the forms of the
empirical syntheses through which such objects must be given.

4. Transcendental Synthesis as an Intrinsically Normative Power of Mind


We have just seen in outline how Kants doctrine of the mental process of synthesis is supposed to produce the characteristic results of his transcendental philosophy of nature. Now, however, the question of normativity confronts us in a
more specific and pressing form: Given that all the schematic structures I have
described are nothing but abstract patterns in the actual processes of empirical
synthesis in the mind, what makes them normative rules for cognition, rather than
merely descriptive generalizations about cognitive psychology? There are two
issues to confront. There is of course the particular problem of what rules for
synthesis Kant proposes, and how these might function as normative rules, capable of answering justificatory questions about our cognitions. Before we even
reach that question, however, there is a more general issue about how Kant could
avoid the psychologistic fallacy at all, if he thinks that any operation of mind, no
matter how described, figures crucially in explanations of the normative standing
of cognition.
We can make progress on the broader question by applying an idea developed
by Gary Hatfield in a more general context.32 Hatfield shows that the charge of
psychologistic fallacy misses the mark, when it is deployed against philosophers
in the early modern tradition, because it trades on an anachronistic conception of
mind. Today, we conceive of the mind as simply a natural feature of human
organisms. Someone like Descartes, by contrast, thinks of the mind as a site of
intrinsically normative powers of knowing, carrying their own standards of
correctness, specifying right and wrong uses, such that when the mind is used
rightly, it will automatically track the truth. Nor is this conception of mind unique
to Descartes, or even to the rationalist tradition. Locke, too, treats the mind as a
normative power, admitting of right or wrong use, and tracking the truth when
rightly used.33 The difference between the two is not over whether the intellect is
a normative knowing power, but over the legitimate scope of that power: Descartes
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

289

thinks we can use it to grasp the real essences of independent substances by intellectual intuition, where Locke thinks we are limited to apprehending the relations
of our ideas.34 Within this general viewpoint, inference from claims about mental
processes to normative results is simply not a fallacy, for the normativity is
already built into the working of the mind.
The idea that powers of mind were supposed to be intrinsically normative is
not often appreciated, probably because it is so foreign to current thinking about
the mind. What intrinsic normativity amounts to may become clearer from a less
philosophically charged example. A jury summons, for instance, has standards of
correctness, rights and wrongs supposed tos built into its being the thing
that it is. When one receives the summons, one is supposed to do certain things:
report to the court at a certain time, bring the summons, park only in certain
areas, notify ones employer, make oneself available during a definite period, etc.
These supposed tos and their normative force are what makes a particular piece
of paper be a jury summons, and not just another letter detailing the services of
your local government. Only something that creates such obligations counts as a
jury summons.
On the early modern conception, normativity is internal to any adequate
description of the mind, just as it is for the jury summons, because the mind is
thought of as an instrument with a correct (intended) use built in. By contrast,
on the currently standard way of thinking, activities and products of mind are
typically evaluated on the basis of standards applied from an outside viewpoint.
So conceived, norms are mind-independent achievements of culture, binding on
particular minds in virtue of their participation in that culture, or in virtue of the
norms objectivity, or what have you; they are not binding in virtue of mindedness as such. We may hope that our processes of belief formation lead to truth,
that our patterns of action lead to goodness and justice, and so on, but these are
independent hopes and desires learned from the social context, and believing in
such norms is not essential to our having a psychology at all.35 Our valuing truth
(or goodness, or validity) does not determine the laws of psychology, nor do
such laws determine the standards of truth (goodness, validity). In this way,
being a mind, in our sense, is something quite different from being a Cartesian
intellect, which comes ready-made from the creator with intended right and
wrong uses.36
This early modern conception of the mind provides the context for understanding Kants question about how knowledge is possible. It even helps explain
why Kant thought the question was so important and revolutionary. He follows
the long-accepted paradigm that treated the mind as a site of intrinsically normative cognitive capacities, but his work is paradigm shattering, in the sense that he
sees this basic assumption as standing in need of special explanation: an adequate
philosophical theory of cognition must explain in detail how it is possible for the
mind to operate as a normative knowing power.37 Kants ultimate, idealist explanation was highly controversial, and I will return to it in the next section. We can
already see, however, that Kant departs from Descartes and Locke not by denying that the intellect is a normative power, but by insisting that they are both
Blackwell Publishers Ltd. 2001

290

R. Lanier Anderson

wrong about which normative powers the mind has (or is). He famously rejects
the very possibility of rationalist intellectual intuition,38 and he is equally sure
that we are not limited to the mere comparison of ideas, as empiricists had
claimed.39
As will be clear by now, Kant took synthesis, instead of intellectual intuition of
essences or mere apprehension of relations of ideas, to be the core normative
power of mind, which is why synthesis plays such a central role in the Analytics
theory of cognition. Above, I raised the possibility that claims about synthesis
might legitimately have normative implications because the rules for synthesis
might receive a normative interpretation. We can now see what this might come
to: the synthesis rules might be rules for the correct use of the (intrinsically
normative) cognitive powers of mind.
But are the rules of synthesis as Kant describes them in fact normative rules?
As we saw above, it is a necessary condition on such rules that they do not entail
their instances, and so exhibit the proper direction of fit. Conceptual rules for
synthesis can be formulated to meet this condition. It would then be possible for
syntheses to violate a given rule, but such syntheses would not count as instances
of the concept. This is not yet sufficient for synthesis rules to be genuinely normative, however, for the direction of fit condition does not yet characterize them as
rules that actually bind us, rules we have reason to follow. For example, I am not
required to synthesize the things all over my desk under the concept <papers>,
rather than under the concepts <notes about Kant>, or even <matter>.40 I must do
so only if I wish to count as using the concept <papers>.41 For all we have seen so
far, such hypothetical requirements, might even be merely non-standard expressions of underlying, essentially descriptive, empirical regularities. In order to
count as part of a water molecule, an oxygen atom must be bound to two hydrogen atoms, but the underlying principle is not normative.
Thus, the fuller sense in which conceptual rules are normative must take us
beyond the hypothetical requirement that only certain syntheses count as
instances of the concept, to a demand that I ought to synthesize using one concept
rather than another. This presents something like genuine cognitive normativity,
by offering a claim that some concept is the right one to use, given the matter to
be synthesized, or other concepts already in play, or both.
The doctrine of a priori synthesis affords rules for synthesis that are normative
in this more robust sense. Consider again the example of perceiving an elliptical
table top. In that case, deploying the concept <table> is optional in just the way
we saw above with <papers>; I might synthesize the perceptual matter using
<table>, but I could use <wood> (or <furniture>, or <ellipse>) instead, or in addition. Note, however, that deploying <ellipse> is not optional in the same way the
others are; there is a sense in which it is required. If I synthesized the perceptual
matter of the table top so that there were no abstract structure implicit in the
synthesis that fit the concept <ellipse> (e.g., if I synthesized such that the area was
not times the product of the radii), then I would have made a mistake in synthesis. The mathematical features of the ellipse are binding for my synthesis in a way
not immediately attainable by empirical concepts. Thus, the structures of a priori
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

291

synthesis serve as normative rules for the syntheses that make up experience;
cognitions are correct when they conform to these rules.
A cognition conforms to the rules of a priori synthesis in a way suggested by
the earlier discussion of synthesis. All syntheses following a given rule for
combining input representations are instances of the same concept. We saw that
concepts can characterize structures at different levels of abstraction in the same
act of synthesis, and a parallel point applies to separate acts of synthesis.
Consider, for example, the syntheses involved in the perceptions of a blue plate
and a red plate: they fail to share the synthetic pattern captured by the concept
<blue>, but do share a more abstract pattern captured by the concept <colored>.
Moreover, as Kant notes in the Schematism chapter (A 137/B 176), they may
also share the much more abstract, indeed a priori, pattern of <circle>. Syntheses
can thus fall under a common concept at one level of abstraction (<circle>,
<colored>), but fail to fall jointly under a less abstract concept (<red>, <blue>). So
syntheses instantiate the same conceptual rule (conform to the concept) just in
case they have a common structure at the relevant level of abstraction that is, if
they are structurally homomorphic. For example, a correct perceptual synthesis of
our table top is homomorphic in this way to <ellipse>, and to <quantity>. Since
categories are conceptual rules, it follows that a given cognition conforms to a
category (and thus satisfies the most basic cognitive norms), if it has an abstract
structure homomorphic to the correct (categorial) form of a priori synthesis. If an
empirical synthesis fails to match the a priori forms, then it violates the norms of
cognition, and is therefore mistaken.
Error, of course, raises a manifest difficulty here. According to the same
synthesis picture, every action of synthesis has more and less abstract structural
features, but obviously, not every synthesis satisfies the norms of cognition. How
does Kant conclude that only some subset of the possible abstract structures for
synthesis comes to count as the class of correct norms for synthesis, which we
have reason to follow? And how can we identify this privileged class?
The short answer is that the normative rules of a priori synthesis are rules of
unity, or coherence, for experience.42 Kants underlying thought rests on deep
considerations about objectivity that make a degree of unity a precondition of the
determinacy of experience. Consider an apparent failure of unity, e.g., two experiences which seem to conflict with one another. From the conflict, we can glean
that they represent either different things altogether, or different states of some
thing which has undergone a change. If we had independent grounds for thinking that the experiences captured different states of one object, then we could
decide in favor of the latter possibility; otherwise, however, the pair of experiences simply fails to represent either possibility determinately. Thus, the pair
becomes a determinate representation only when the two are unified by being
treated as representations of one (possibly changing) object.43 Of course, sometimes we have experiences that do simply represent different things, but note
that they cannot determinately and explicitly represent their respective stretches
of experience as different (rather than a single, changing thing), except by placing
the things they represent in a definite relation to one another in one underlying
Blackwell Publishers Ltd. 2001

292

R. Lanier Anderson

collection of objects. Even in this instance, then, determinate representation


depends on the kind of underlying unity we achieve when we successfully
attribute different representations to an object; here we just confront the limit case
in which the one object is nature itself, and different things are related to one
another as its parts.44
Kant concludes that the norms establishing the proper use of the cognitive
faculties should be rules of unity, for it is only by unifying our representations in
this way that we can use them to achieve determinate, objective representation in
the first place. Therefore the correct forms of synthesis (i.e., the genuinely a priori
ones) will be those whose structure fits to a very large amount of our experience,
and so enables us to unify that experience into a coherent picture of the world.
This is why Kant identifies the form of experience by applying the regressive
method to cases of especially systematic cognition, like mathematics and exact
natural science. The abstract features of synthesis in these cases are likely to
produce highly general structures of synthesis, with which our various syntheses
can best be reconciled with one another into a maximally unified whole.
Empirical syntheses will be correctly carried out when their structure is homomorphic to these a priori forms. When Kant argues that a particular form of
synthesis rises to the level of a condition of the possibility of experience, he is
trying to guarantee its status as a principle of the unity of experience, which is
thereby normative for empirical synthesis.
It does not follow, of course, that no empirical synthesis could ever conflict
with the general, a priori structures of synthesis. Naturally, some actual representations will not fit into our objective picture of the world: we dream, have perceptual illusions, make false judgements, etc. In these cases, something about the
synthesis resists homomorphism with the general, unifying forms. Suppose, for
example, that I dream I have missed my friends wedding because I was caught
up watching reruns, and that she is now angrily slamming a door in my face
and then I wake up disoriented in my hotel room to the sound of jackhammers
ripping up the street outside. My dream experiences cannot be integrated with
nearby experiences. (The door disappears; its sound is replaced by the jackhammers; my friend is not really angry with me; it is still two days before the
wedding; etc.) Such deviant representations fail to match the general norms of
synthesis (in this case norms associated with the categories of cause and
substance), and so they are evaluated as non-veridical, or, as Kant often puts it,
merely subjective (A 201/B 247).45 In the same way, we would dismiss anyone
who calculated the area of our elliptical table top by some other rule than times
the product of the radii. That structural rule has normative force for our experience, and any experience that seems to conflict with it will be discounted.
The doctrine of a priori synthesis is thus crucial to Kants explanation of how
cognition is possible, because he thinks the a priori level of synthesis provides the
fundamental cognitive rules for the minds functioning as a normative knowing
power. The a priori status of transcendental synthesis does important work here.
First, it allows synthesis rules to satisfy the necessary direction of fit condition.46 Natural laws imply their instances, so violations disconfirm the law, but
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

293

normative rules remain binding even when violated, and we hold the particular
events (e.g., false judgments, bad actions) accountable to them. Since a priori
synthesis is not dictated by the matter given through sensation, it has standing
independent of how the details of experience are filled in. Therefore, the structures of a priori synthesis are not caused by or accountable to experience; they are
not merely a links in a causal chain of belief formation processes governed by
exceptionless Humean laws of association; and, qua a priori, they have the appropriate direction of fit to serve as normative rules, against which the empirical
syntheses are to be evaluated. Second, the a priori synthesis provides the form for
veridical experience, which allows empirical syntheses to count as determinate
representations of cognitive objects in the first place. We therefore have reason to
synthesize under these rules, and their structure normatively constrains all filling in by sensory matter, or content. The distinction between the a priori and
empirical levels of synthesis thus gives Kant the resources to make a distinction
between valid and invalid instances of synthesis, and thus to count synthesis as a
normative power of the mind, which has right and wrong uses intrinsically.

5. Explaining How Knowledge is Possible: the Import of Transcendental


Idealism
Even if we agree for the moment to put aside worries about Kants unfamiliar
normative conception of mind, the picture of a priori synthesis just sketched still
raises a fundamental question: Why should objects conform to whatever representations our rules of synthesis require us to produce? Put another way, the independent standing of the a priori structure within synthesis might enable it to serve
as some norm against which we might decide, for whatever reason, to evaluate
our various empirical syntheses, but why should we think that the norm to which
the rules of a priori synthesis tunes our representations is the norm of truth?
At this point, the full meaning of Kants question about how knowledge is
possible comes into focus. The possibility of knowledge is a special problem
because knowledge is a normative achievement, and this achievement actually
connects us to the world. According to Kant, a non-arbitrary connection between
cognition and the world can be established only if there is a real influence in one
direction or the other: either the world makes our concepts be the way they are,
or our concepts make the world be the way it is (B 1668, B xvixviii). Kant thinks
all attempts to make out the first alternative are doomed. Empiricist accounts of
belief formation compromise the normative standing of cognitive achievement,
because they provide a mere physiology of the understanding (A ix), i.e., a set
of natural laws of the mind. Rationalist accounts do not eliminate cognitions
normativity by making its connection to the world a matter of natural law, but
their strategy renders the connection itself unintelligible. They turn it into a miracle, by appealing to some magical power (A xiii) of cognition, like intellectual
intuition, which in the end must be underwritten by special sanction from a
benevolent God. Kant notoriously concludes that the only way to guarantee a
Blackwell Publishers Ltd. 2001

294

R. Lanier Anderson

non-arbitrary, but still normative, connection between cognition and the world is
to reverse our usual assumptions, and opt for a system of idealism, in which our
conceptual structures dictate the structure of the natural world.
Kant returns to his insistence on idealism at the close of many of the arguments
of the Transcendental Analytic.47 His claim in these passages is that the valid
applicability he has shown for the category in question, and therefore genuine
knowledge using that category, is possible only because the objects of knowledge
are appearances. That is, the objects are just the fillings in by empirical details of
a priori valid schemata for synthesis. Therefore, the conditions placed on the
proper representation of objects by the normative rules of synthesis are at the
same time conditions on the things to be known.
Thus, idealism was so important to Kant because it is the answer to his question about how knowledge is possible. That turns out to be a question about how
the mind can function as a truth-tracking normative power. This will be possible,
in turn, only if two conditions obtain: 1) the mind must carry within itself an
intrinsic distinction between its right and wrong uses; and 2) the right use must
track the truth about objects. The forms of the transcendental synthesis establish
which uses of the mind are correct, and this meets the first condition. Idealism
meets the second. Under this system, the realm of nature draws its fundamental
structure from just this transcendental synthesis, considered as a kind of metaphysical/mathematical schematic construction that outlines the abstract form of
nature. Because the a priori synthesis constitutes the world of appearance in this
way, empirical syntheses that conform to its form will capture the truth about
that world.

6. Conclusions
Insofar as it means to treat knowledge, then, the meaning of Kants question is to
ask how the mind can be a normative knowing power, and ultimately to ask how
cognitive normativity is possible at all i.e., how we can achieve cognitions
which have a non-arbitrary connection to objects, but are not simply caused by
their objects in a way that would compromise their normative status. Although
greatly concerned with cognitive mechanisms, Kants question is not merely
psychological, in our post-Kantian sense. Nor (despite its concern with justification) is it merely epistemological, since it involves itself deeply in the philosophy
of mind, and in the concrete workings of the cognitive powers.
Kants idealist solution to the question is clearly powerful, especially when it
is linked to his detailed picture of a priori synthesis, to the roots of that picture in
his account of eighteenth century mathematical practice, and to the status, in the
philosophical context, of intrinsically normative powers of mind as a going philosophical theory. Nevertheless, the theoretical costs of Kants approach are bound
to seem very high to us. His early modern conception of mind is no longer a live
philosophical option; mathematical practice has abandoned the proof procedure
he presents as unavoidable; and Kants idealism seems highly problematic. The
Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

295

idealism, in particular, struck even many of his contemporaries as unacceptable. It


must be admitted, however, that idealism offers quite an elegant solution to the
problem of explaining how the mind could function as a normative knowing
power. Given the obviousness of that problem once Kant had posed his question
and the elegance of the idealist resolution, it is possible that idealism came to
seem so necessary to underwrite the normative conception of mind, that the threat
of such idealism itself became one reason for subsequent philosophy to abandon
that conception, and cede the province of mind to naturalistic psychology.
Whatever the details of these historical developments in our conception of the
mind, the Kantian contribution to that history has unquestionably left us with a
significant piece of our current philosophical problematic. Kants effort to pose
the question about the possibility of cognition as a normative achievement
convinced many philosophers of the importance of a fundamental distinction
between the natural and the normative, and without intrinsically normative
mental capacities to serve as the source of normativity, philosophy since the midnineteenth century has spent a great deal of effort searching for a different way
to ground normative practices of culture. The very difficulty of the search
measures the price of relinquishing the admittedly troubling assumptions out of
which Kant erected his solution.48
R. Lanier Anderson
Department of Philosophy
Stanford University
Stanford
CA 943052155
USA
lanier@csli.standord.edu

NOTES
1

Citations to the Critique of Pure Reason use the standard A/B format to refer to the
pages of the first (A) and second (B) editions. Citations to Kants other works employ the
pagination of the Akademie edition. I have used the translations and abbreviations listed
among the references.
2 Kants statements about the questions centrality are already prominent in the A
edition and the Prolegomena. In A Kant wrote:
A certain mystery thus lies hidden here, the elucidation of which alone can make
progress in the boundless field of pure cognition of the understanding secure and
reliable: namely, to uncover the ground of the possibility of synthetic a priori judgments . . . . (A 10)
Complaining about the Garve-Feder review in the Prolegomena, Kant wrote that it:
did not mention a word about the possibility of synthetic cognition a priori, which
was the real problem, on the solution of which the fate of metaphysics wholly
rests, and to which my Critique (just as here my Prolegomena) was entirely directed.
(Prol. 377)
Blackwell Publishers Ltd. 2001

296

R. Lanier Anderson

3 Kitchers work (1990, 1982, 1984, 1995, and 1999) has paved the way for recent
psychological treatments of Kant, reopening an interpretive tradition which had all but
disappeared from view since the nineteenth century. See Brook 1994, for example, and
also, to some extent, Falkenstein 1995. Hatfield 1992 traces the different kinds of psychology present in Kant, and clarifies the use of psychological concepts in the central arguments of the Analytic, but he cannot be simply classed within the psychological reading.
What distinguishes that approach is the idea that the Analytic offers an essentially psychological theory, whereas Hatfield 1992 draws a strong distinction between psychology and
the transcendental philosophy of the Analytic. See also Hatfield 1990.
4 The term is due to F. A. Lange (1876, II: 30), whose chapter on Kant is of special interest in this connection.
5 In many respects, Lange proposes replacing Kants transcendental account with an
empirical theory of the mind. He claims, e.g., that Kants great idea was to discover (Lange,
1876, II: 29) the a priori components of experience, but that he had gone about this the wrong
way: The metaphysician must be able to distinguish the a priori concepts that are permanent and essentially connected with human nature, from those that are perishable, and correspond only to a certain stage of development. . . . But he cannot employ for this other a
priori propositions. . . [or] so-called pure thought, because it is doubtful whether the principles of those have permanent value, or not. We are therefore confined to the usual means
of science in the search for and examination of universal propositions that do not come from
experience; we can advance only probable propositions about this. . . (Lange, 1876, II: 31).
Similar ideas have resurfaced, along with the psychological reading itself, in the twentieth
century. Compare Kitcher 1990: 846, 11112, 1356; 1995: 302, 306 and Brook 1994, who
relate Kants ideas to results of contemporary experimental psychology.
6 Hume claims that all events associated by cause and effect (including occurrences of
ideas in the mind) are loose and separate (Enquiry VII, ii; Hume 1975: 74), and that we
have no legitimate idea of real or necessary connections which might link them. He explicitly applies this skepticism to connections among ideas in the discussion of personal identity in the Appendix to the Treatise: If perceptions are distinct existences, they form a
whole only by being connected together. But no connexions among distinct existences are
ever discoverable by human understanding, because the mind never perceives any real
connexion among distinct existences (Hume 1978: 635, 636). For Kants claim to the contrary
that the understanding does forge real connections among perceptions, see the beginning
of the Second Analogy, at B 233, and also A 225/B 272.
7 See, e.g., criticisms of Kants psychology by Strawson 1966, and Guyer 1987 and
1989. Guyer 1989 thinks that it is possible to find a non-psychologistic theory of synthesis
in the Analytic, which proceeds by identifying conceptual truths about any representing
or cognitive systems. . . that work in time (58). This approach does make claims about the
mind, but Guyer thinks it avoids postulating particular mental acts, and therefore dependence on contingent empirical psychological truths. As such, it does not fall into psychologism. The view I offer below is indebted to Guyers treatment (see also Guyer 1992a).
8 For Kants understanding of empirical psychology, see Prol., 295, A 347/B 405, A
100, A 1123, A 122, and B 152. In outlining his compatibilism, too, Kant subjects the empirical character to exceptionless natural laws (A 539/ B 567, A 545/B 573, A 54950/B 5778).
For the separation between Kants transcendental theory of cognition and the empirical
physiology of inner sense, see A 85/B 117, A 867/B 1189, and also A 545/B 789,
where Kant separates empirical psychology from pure logic, which contains transcendental logic, and hence the theory of cognition of the Analytic. Detailed analysis of Kants position on psychology can be found in Hatfield 1992, esp. pp. 21724.

Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

297

9 Kants anti-psychologistic treatment of general logic (A 535/B 779) emphasizes


that an empirical account of the minds operations cannot explain the normative force of
logical laws.
10 The whole aim of such causal stories is to explain the cognitive states that are actually produced in the event, by deriving them from the prior facts and the laws.
11 To my knowledge, the first to make this point as an attack on the psychological reading was Hermann Cohen 1871.
12 For example, in the Transcendental Deduction Kant writes that:
Jurists, when they speak of entitlements and claims, distinguish in a legal matter
between questions about what is lawful (quid juris) and that which concerns the fact
(quid facti), and since they demand proof of both, they call the first, that which is to
establish the entitlement or the legal claim, the deduction. . . . Among the many
concepts that constitute the mixed fabric of human cognition, there are some that are
also destined for a pure use a priori, and these always require a deduction of their
entitlement, since proofs from experience are not sufficient for the lawfulness of
such a use. . . . [They require] an entirely different birth certificate than that of ancestry from experience. I will therefore call this attempted physiological derivation,
which cannot be called a deduction at all because it concerns a quaestio facti, the
explanation of the possession of a pure cognition. (A 847/B 11619)
13 Paul Guyer is perhaps the leading contemporary exponent of this approach (see
Guyer 1987, esp. pp. 232, 2416, 3035, 31516, and Guyer 1989), but such epistemological
readings have been dominant since Cohen 1871, and most recent commentaries fall into
this broad camp (including, e.g., Allison 1983, with its central focus on the idea of an epistemic condition). Wolff 1963 is an exception, as are the contemporary psychological readings cited above.
14 Kant himself calls geometry a cognition [eine Erkenntnis] (Prol. 280; cf. also A 87/B
120, et passim), and so for him the validity of geometry is arguably not a wholly mind-independent fact, but a fact about human cognitive achievement. Since, however, this cognition is, by Kants own lights, wholly independent from any particular mind, his
anti-psychologistic followers can easily eliminate any vestiges of mentalism from Kants
formulations, by construing geometry as a body of doctrine, which is what it is no matter
what human beings know about it. Indeed, it is arguable that Kant never meant anything
else in referring to geometry as eine Erkenntnis; such a view is implicit, e.g., in Hatfields
choice to translate the phrase by a body of cognition.
15 Versions of the broadly anti-psychologistic approach have been advanced by scholars with otherwise very widely diverse positions, including Cohen 1871, Windelband 1884,
Cassirer 1981 [1918], Kemp Smith 1923, Strawson 1966, Bennett 1966, Pippin 1982, Allison
1983, and Guyer 1987 and 1992a.
16 Strawson offers a particularly striking example of the tendency:
Kant. . . thought of his project in terms of a certain misleading analogy. . . . the character of our experience. . . is partly determined by our cognitive constitution. . . .
[But] the workings of the human perceptual mechanism. . . are matters for scientific, not
philosophical investigation. Kant was well aware of this. . . . Yet, in spite of this awareness,
he conceived the latter investigation by a kind of strained analogy with the former.
Wherever he found limiting or necessary general features of experience, he declared
their source to lie in our own cognitive constitution, and this doctrine he considered
indispensable . . . Yet there is no doubt that this doctrine is incoherent in itself, and masks,
rather than explains, the real character of his inquiry; so that the central problem in
understanding the Critique is precisely that of disentangling all that hangs on this

Blackwell Publishers Ltd. 2001

298

R. Lanier Anderson

doctrine from the analytical argument which is in fact independent of it. (Strawson
1966: 1516, emphasis added).
17 Wayne Waxman (1991: 250n28, 25960) rightly points out the need for care with the
notion of synthesis. Kant sometimes clearly uses synthesis to refer to the combining activity alone (i.e., the second feature in the list), whereas at other times synthesis refers to a
more complex achievement involving all three aspects. Waxman is correct that for certain
arguments, Kant actually needs a notion of synthesis that does not analytically contain the
third element (the idea of unity), which would beg certain questions against Hume, Tetens,
et al., by building into the very concept of mental processing a strong form of unity that
Kant needs to show belongs to our cognitions. For these reasons, Waxman prefers to restrict
Kants uses of synthesis to the former meaning. (He strictly distinguishes synthesis from
combination, which involves all three elements, and he criticizes scholars (e.g., Kitcher 1990:
737) who simply equate the two.)
I think, however, that the clear distinction is due to Waxman, and not Kant. Kant
himself clearly equates synthesis and combination at B 12930 that is, right at the
beginning of the B Deduction, where he should be most sensitive about begging the question against empiricism. Moreover, there is a clear use of synthesis in the more restricted
sense Waxman prefers on the very same page (B 130). I think it better to conclude that Kant
uses synthesis to mean both mere combining (i.e., synthesis sensu Waxman), and also the
more complex achievement that involves such combining, plus the ideas of the manifold
and of a unity guiding combination. This usage makes sense for Kant, because he ultimately aims to argue that even the lower level, more basic kinds of synthesis in fact
depend on higher level, conceptualized forms of synthesis.
18 Thus Kant: all combination, whether we are conscious of it or not,. . . is an action of
the understanding, which we would designate with the general title synthesis (B 130).
19 Kant often speaks of functions of synthesis (see, e.g., A 105, A 108, A 109, A 112, A
164/B 205, A 181/ B 224), and he also says that concepts rest on functions (A 68/B 93).
Kant defines function as the unity of the action of ordering different representations
under a common one (A 68/B 93). Thus, concepts rest on functions in the sense that they
provide the representation of unity that belongs to and guides a synthesis, so that the
unified synthesis can be abstractly captured by a kind of function, or mapping together,
that links representations.
20 This account is indebted to Kitcher 1990, who also sees syntheses as functions generating a contentual connection (117) that establishes an existential dependence among
representations (103, 117). But not every existential dependence relation among contents
amounts to Kantian synthesis. Suppose I see a picture of my friend Peter, who is in France,
and this gives me the idea of Peter (via resemblance, the second Humean law of association). The second representation is existentially dependent on the first: the idea exists in
me because the first (visual) representation occurred. Is the second also contentually dependent on the first? In a sense it is, but in a sense not, and the difference brings out Kants
departure from a merely causal theory of association. It is true that the second is an idea of
Peter because the first was an impression of a picture resembling him. But the Humean
account does not envision that the content of my idea of Peter is altered by the association:
association effects a mere co-location in the mind of the two representations, without
necessarily altering the ideas content. Kantian synthesis, by contrast, pulls apart elements
of the contents of the inputs and reorders them in the output representation, which is why
synthesis is essentially conceptual, not merely causal combination. The reordering of
inputs by synthesis follows a pattern provided by a concept, not order that arises automatically from the partial contents themselves according to laws. Otherwise, we still have

Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

299

only a case of Humean association (now operating among the simpler parts of the representations). What makes conceptually driven synthesis distinctive, then, is that combination is dictated at the level of the contents, for which no particular causal realization of the
process would matter. (Two syntheses could be instances of the same synthetic pattern
even if they were realized in processes where the contents were carried by different kinds
of objects, which would fall under different causal laws.)
21 Waxman 1991 offers such a bottom up approach to the theory of cognition, as does
Kitcher (see her 1990 and 1995), though my account is nevertheless indebted to hers on
many details. Perhaps the most interesting recent interpretation on this general question
of approach is that of Longuenesse 1998, whose reading combines elements of both top
down and bottom up approaches by giving the categories two different, essential roles
in the process of cognition, one at each end of the procedure of constructing empirical
knowledge.
22 See Longuenesse 1998 for a provocative account of Kants guiding thread idea.
23 Thus, both the abstract concepts of the understanding and the concrete data of sense
contribute essentially to the resulting perception. Some features of the table (e.g., brown
color) are conferred by sense, and without such detailed determinations, the conceptual
structures of the understanding would have no objective reality. So the categories do not
fully determine the empirical content that fills in experience, but they nevertheless constrain
the broad shape taken by that content, whatever the empirical details turn out to be.
24 The locus classicus for this traditional conception of a priori synthesis (and for criticism
of Kants position) is Bennett 1966: 11117, who assumes that, if it is to be a priori, transcendental synthesis must somehow occur outside of time. Since no action like synthesis could
happen outside time, the whole doctrine seems desperately unpromising (111). Kitcher
1982: 539 disagrees, treating transcendental syntheses as a special subclass of empirical
syntheses, known to exist in a special way (viz. as conditions of the possibility of experience). Her view, however, still leaves transcendental syntheses as separate acts of the mind,
rather than making such synthesis an aspect of all (veridical) empirical syntheses.
The conception of a priori synthesis defended here owes the most to Guyer. While he
also sometimes talks about the a priori synthesis as a separate action of the mind from
empirical syntheses (Guyer 1980: 205, 2067, 208; 1987: 136), the account of synthesis in
Guyer 1989 (see note 8, above) offers a reading of the relation between a priori and empirical elements of cognition similar to the one I propose. (N.B., Guyer 1980 also shows
convincingly that Kants a priori synthesis must be understood as an action of the mind, not
as an unfortunate way of expressing a point about a priori criteria for knowledge, as
Bennett would have preferred Kant to mean.)
25 For extended treatment of that distinction, see Pippin 1982.
26 Kant insists that mathematical proof depends on ostensive construction to reach its
results (see Friedman 1985, 1992a; Shabel 1997, 1998; and for an earlier version of the point,
Philip Kitcher 1975). Kants view is a good account of eighteenth century textbook mathematics (see esp. Shabel 1998). In particular, the essential reliance on the diagram in
Euclidean proofs provides strong evidence for Kants views on construction, and as Shabel
1997 shows, the textbooks Kant used in teaching his mathematics courses treated this
geometrical procedure as fundamental to mathematical science, presenting arithmetic and
algebra as similarly dependent on construction.
27 Note here that angle size is not a feature abstracted from by this schema in the same
way as side size. For we can extend the present construction to show the equality of the
three angles (by iterating the construction strategy deployed in Euclids Bk. I, Prop. 5,
proving the equality of the angles formed by the base of an isosceles triangle). It follows
Blackwell Publishers Ltd. 2001

300

R. Lanier Anderson

that the angles of an equilateral triangle are equal to one another, and no instantiation that
lacks this property will count as a proper filling in of the schema provided by the construction of Euclid I, 1. The key implication is that the schema carries intuitive information, in
Kants sense. That is, the property about angle size is dictated by the schema, but only
because of proofs which appeal essentially to intuition, and would not be possible on the
basis of concepts alone.
28 The structural features already expressed in the schema will also be the ones
exploited by the synthetic a priori proofs that rely on the construction in question. For
example, Lisa Shabel has persuasively argued (in her 1997, 1998) that geometrical
constructions carry information about relative magnitudes, by virtue of explicitly representing gross part/whole relations among features of the constructed figure. For a
summary of some details relevant to Shabels results, see also Anderson (unpublished
manuscript-a).
29 For a discussion of the importance of the Schematism with important similarities to
the one offered here, see Rosenberg (1997), which focuses especially on the way the
doctrine of schematism is supposed to solve the problem of the unity of perception (i.e.,
how the presentational, or phenomenal, content of a perception is combined with its cognitive, conceptual, or propositional content as a cognition of a particular thing, under a given
concept).
30 This is, of course, the task of Kants Schematism chapter (A 13747/ B 17687).
31 See Friedman 1992a (esp. chs. 3 and 4), 1992b, and 1993.
32 See Hatfield 1997, and also Hatfield 1990.
33 The case for attributing this conception to Locke might not seem quite as strong to
some as the parallel case for Descartes. But consider the following remarks from Lockes
Essay (following Locke 1975):
[Some ideas offer themselves to us more readily than others,] Though that too, be
according as the Organs of our Bodies, and Powers of our Minds, happen to be
employd; God having fitted Men with faculties and means, to discover, receive, and retain
Truths, accordingly as they are employd. The great difference that is to be found in the
Notions of Mankind, is, from the difference they put their Faculties to, whilst some . . .
misimploy their power of Assent, . . . Others . . . attain great degrees of knowledge . . . .
(I, iv, 22; emphasis in original)
Here Locke clearly envisions Powers of our Minds, which have a right and wrong use
that is proper to them, and which produce knowledge (only) when rightly used. Or again:
so far as this faculty [of accurately discriminating ideas] is in it self dull, or is not rightly
made use of, for the distinguishing of one thing from another; so far our Notions are
confused, and our Reason and Judgment disturbed or misled. (II, xi, 2)
34 Even Hume clearly recognized such powers of the mind to apprehend relations of
ideas by the time of the Enquiry (IV, i; Hume 1975: 2532).
35 For a clear elaboration of this view, see Dretske 2000, which claims that norms come
from us from our intentions, purposes, desires (Dretske 2000: 245). This means they
come from particular (contingently present) mental formations, not from being a mind as
such. Even a philosopher like Brandom (1994), who argues (against the current grain) that
normativity is essential to certain aspects of our mental life (language use, full intentionality), still takes norms to be fundamentally social, not intrinsic to individual mindedness.
Thus, for Brandom, animals have psychological lives and share responsive capacities analogous to our lower cognitive processes, but they possess only derivative intentionality
precisely because they do not operate in the right kinds of normative social practices. Fully
intentional Brandomian minds are perhaps inherently susceptible to be trained to become

Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

301

responsive to norms, but the norms are in the training; they are not already there in the
mind, built in, waiting to be revealed by Cartesian cognitive meditation. That even
Brandom is so naturalistic in this respect is one measure of how far we have departed from
the early modern conception of intrinsically normative powers of mind.
36 To regain a feel for the difference, recall that training people in these right and
wrong uses is the goal of Descartess keen insistence that we meditate with him. Descartes
knows that it will be difficult for us to attain clear and distinct perceptions; he designs his
text so carefully because he wants to get us to experience the correct use of the intellect, i.e.,
its use freed from the distractions of the senses. On the importance of seeing the
Meditations within the tradition of spiritual training exercises, see Hatfield 1986, Kosman
1986 and A. Rorty 1986.
37 The need to explain the mechanisms of normative cognition comes into sharper
focus due to Kants separation of empirical psychology from the transcendental theory of
cognition. Empirical psychology treats mental events in independence from their normative properties, so in light of Kants distinction, it no longer goes without saying that they
can be intrinsically normative. That now needs explanation. Empiricists (especially Hume)
also often adopted a psychological perspective on the mind, but did not similarly contrast
it against the minds normative use in cognition. As a result, their explanations of knowledge focus either on topics within psychology (which do not address cognitions normativity, from Kants viewpoint), or on arguments that the mind cannot acquire certain kinds
of knowledge. Hatfield (1997: 30) plausibly suggests that Locke saw no need to explain the
minds normative cognitive power, since his aim was to deny that it could grasp real
essences, in any case. On the positive side, Locke rests content with the visual metaphor
that the mind simply perceives connections of ideas (e.g., the visible connections of IV, iii,
14), without analysis of such perception beyond appeals to introspection (see IV, i, 24;
IV, iii, 14).
It could be argued that rationalist philosophers, who made more extensive claims for
the intellect, did try to explain its achievements, e.g., by appealing to a divinely guaranteed power of intellectual intuition, or to various forms of divinely established harmony
between our ideas and essences resident in Gods intellect. From Kants point of view,
however, such accounts lack real explanatory power, particularly when it comes to the
mechanics of the connection between knowledge and its objects. As Locke already points
out (IV, iii, 6), we explain connections between mind and world via divine intervention
precisely when we no longer understand how they are possible. For Kants part, pure intellectual intuition counts among the magical powers (A xiii), and any preformation system
of pure reason (B 1678) forsakes explanation of the mechanisms of the normative connection involved in knowledge, since it makes that connection accidental, from the point of
view of the cognition and its object themselves. Comments from an anonymous reviewer
for EJP helped clarify my thinking on these issues.
38 See B 68, B 712, A 51/B 75, B 135, B 1389, B 145, and as mentioned in the previous
note, the A Preface dismissal of magical powers capable of satisfying the dogmatically
enthusiastic lust for knowledge (A xiii) in metaphysics.
39 As Kant wrote in the opening paragraphs of the B edition: There is no doubt whatever
that all our cognition begins with experience. . . . But although all our cognition commences
with experience, yet it does not on that account all arise from experience. For it could well be
that even our experiential cognition is a composite of that which we receive through impressions and that which our own cognitive faculty . . . provides out of itself (B 1).
40 Angle brackets (< >) indicate the mention of a concept.
41 Cf. a related point about Kants argument in the A Deduction in Guyer (1987: 1067).
Blackwell Publishers Ltd. 2001

302

R. Lanier Anderson

42 This idea is ubiquitous in the Critique: for example, Kant writes that Experience
therefore has principles of its form which ground it a priori, namely general rules of unity
in the synthesis of appearances (A 157/B 196), and that the empirical truth of appearances
is satisfactorily secured, and sufficiently distinguished from its kinship with dreams, if
both [space and time] are correctly and thoroughly connected up according to empirical
laws in one experience (A 492/B 5201). Cf. also A 177/B 220, A 181/B 2234, A 216/B
263, B 2789, A 228/B 281, A 22930/B 282, and A 4945/B 523. The texts at B 2789 and
A 492/B 5201 explicitly emphasize that syntheses failing to conform to the rules of unity
for experience are non-veridical.
43 This idea is related to the deep thought behind Kants theory of time-determination
in the Analytic, which claims that without objective rules for time-ordering experiences, it
is not possible to represent a difference between a representation of a change in one thing,
and a representation of two separate things. For a sustained account of the theory of timedetermination see Guyer 1987: 207329.
44 Kants Transcendental Deduction of the Categories argues that the unity in the
content of experience discussed in this paragraph (the unity of the world represented) is
just the objective correlate of the unity of the consciousness of the representing subject (the
unity of apperception). The argument there attempts to exploit our knowledge of the unity
of apperception on the subject side to show the validity of the categories necessary to bring
about such unity in our representation of the world, on the object side.
45 Thus, Kant writes that if the causal law were apparently violated in some representation, then I would have to hold it to be only a subjective play of my imaginings, and if I
still represented something by it I would have to call it a mere dream (A 2012/B 247); or
again, it does not follow that every intuitive representation of outer things includes. . .
their existence, for that may well be the product of imagination (in dreams as well as delusions). . . . Whether this or that putative experience is not mere imagination must be ascertained according to its particular determinations and through its coherence with the
criteria of all actual experience (B 2789).
We can handle perceptual illusions in essentially the same way as dreams. When I see
a stick in water, and it appears bent, the content of the experience resists integration with
other, nearby experiences (e.g., perception by touch of the sticks straightness, the perception of the straight stick being pulled out of the water). Of course, I can connect the experience to others by causal laws (by reverting to laws of optics and sensory psychology),
but such connections require me to treat the representations as themselves objects covered
by the relevant causes, rather than applying causal interpretation to the apparent content
of the representations. Precisely this is what marks them as subjective in Kants sense.
They belong to my subjective unity of consciousness (B 139), which is valid only for me
precisely because the causal story connecting its constituents depends on contingent
initial conditions of my psychological situation. Nevertheless, even this subjective unity
is derived from the former [objective unity] under given conditions in concreto (B 140),
in the sense that the causal laws of optics and psychology that explain the illusion are
objective, and bring the representations, considered now as objects, into the original
unity of consciousness, or transcendental unity of apperception (B 140, 139). Thanks to
Alison Simmons, Daniel Sutherland, and Andrew Janiak for pressing me to clarify this
point.
46 On the basis of reasons like those suggested in the text, in fact, Kant seems to have
been gripped by the intuition that the binding force of any and every norm would have to
be explained by some a priori ground. In the case of moral philosophy, for instance, Kant
writes that:

Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

303

These principles [of morality] must have an origin entirely and completely a priori, and
must at the same time derive from this their sovereign authority. . . . Hence everything
that is empirical is, as a contribution to the principle of morality, . . . highly injurious to
the purity of morals; for in morals the proper worth of an absolutely good will. . . lies
precisely in this that the principle of action is free from all influence by contingent
grounds. . . . (Groundwork 426; my emphasis)
I discuss the point in detail, and raise difficulties for Kants view, in Anderson (manuscript-b).
47 Kant gives the transcendental idealist upshot of his theory of cognition especially
prominent play at the end of the Transcendental Deduction in both editions (see A 12930
and B 1635), but he makes essentially the same point repeatedly throughout the Analytic.
See, for example, A 13940/B 1789 and A 1467/B 1867 in the Schematism chapter; A
166/B 2067 in the Axioms; A 1801/B 2234 at the end of the general introduction to the
Analogies; a number of places in the Postulates of Empirical Thought (A 21920/B 2667,
A 224/B 272, A 2278/B 280); and, prominently, the General Note to the System of
Principles (B 289, 294). There is also, of course, the general discussion of related issues in
the Analytics closing chapter on Phenomena and Noumena.
48 I am grateful to Kritika Yegnashankaran for the invitation to give the talk that
resulted in this paper. Thanks to Lori Gruen, Paul Guyer, Gary Hatfield, Vittorio Hsle,
Andrew Janiak, Stephan Kufer, Joshua Landy, Elijah Millgram, Katherine Preston, Tim
Schroeder, Alison Simmons, Daniel Sutherland, and Allen Wood for helpful comments
and conversations about earlier versions, and to audiences at Stanford University, San Jose
State University, California Polytechnic Institute (San Luis Obispo), Arizona State
University, the University of Miami, and the Ninth International Kant Congress for useful
comments and questions.

REFERENCES
Allison, Henry (1983), Kants Transcendental Idealism. New Haven, CT: Yale University
Press.
Anderson, R. Lanier (unpublished manuscript-a), Contextualizing Kants Philosophy of
Mathematics, Stanford University.
- (manuscript-b), Normativity, Psychologism, and the Human Sciences: an
Investigation into the Motivations of Early Neo-Kantianism, Stanford University.
Brandom, Robert (1994), Making it Explicit: Reasoning, Representing, and Discursive
Commitment. Cambridge, MA: Harvard University Press.
Brook, Andrew (1994), Kant and the Mind. Cambridge: Cambridge University Press.
Bennett, Jonathan (1966), Kants Analytic. Cambridge: Cambridge University Press.
Cassirer, Ernst (1981 [1918]), Kants Life and Thought. Trans. J. Haden. New Haven, CT: Yale
University Press.
Cohen, Hermann (1871), Kants Theorie der Erfahrung. Berlin: Ferd. Dmmlers
Verlagsbuchhandlung, Harrwitz und Gossmann. Reprinted as Kants Theorie der
Erfahrung (1871). Hildesheim: Georg Olms Verlag, 1987. (Band I.3 of Cohens Werke.)
Dretske, Fred (2000), Norms, History, and the Constitution of the Mental, in his
Perception, Knowledge, and Belief. Cambridge: Cambridge University Press, pp. 24257.
Falkenstein, Lorne (1995), Kants Intuitionism: a Commentary on the Transcendental Aesthetic.
Toronto: University of Toronto Press.
Blackwell Publishers Ltd. 2001

304

R. Lanier Anderson

Friedman, Michael (1980) Kants Theory of Geometry, Philosophical Review 94: 455506.
- (1992a), Kant and the Exact Sciences. Cambridge, MA: Harvard University Press.
- (1992b), Causal Laws and the Foundations of Natural Science. In Guyer, ed. 1992b.
- (1993), Kant and the Twentieth Century. In P. Parrini, ed., Kant and Contemporary
Epistemology. The Hague: Kluwer, pp. 2746.
Guyer, Paul (1980) Kant on Apperception and A priori Synthesis, American Philosophical
Quarterly 17: 20512.
- (1987), Kant and the Claims of Knowledge. Cambridge: Cambridge University Press.
- (1989), Psychology and the Transcendental Deduction. In E. Frster, ed. Kants
Transcendental Deductions: the Three Critiques and the Opus Postumum. Stanford, CA:
Stanford University Press.
- (1992a), The Transcendental Deduction of the Categories. In Guyer, ed. 1992b.
-, ed. (1992b) The Cambridge Companion to Kant. Cambridge: Cambridge University
Press.
Hatfield, Gary (1986), The Senses and the Fleshless Eye: the Meditations as Cognitive
Exercises. In Rorty, ed. 1986b.
- (1990), The Natural and the Normative: Theories of Spatial Perception from Kant to
Helmholtz. Cambridge, MA: MIT Press.
- (1992), Empirical, Rational, and Transcendental Psychology: Psychology as Science
and as Philosophy. In Guyer, ed. 1992b.
- (1997), The Workings of the Intellect: Mind and Psychology. In Easton, P., ed., Logic
and the Workings of the Mind: the Logic of Ideas and Faculty Psychology in Early Modern
Philosophy. Atascadero, CA: Ridgeview/North American Kant Society, pp. 2145.
Hume, David (1975), Enquiry Concerning Human Understanding. Ed. L. A. Selby-Bigge, P.
Nidditch (3rd ed.). Oxford: The Clarendon Press.
- (1978), A Treatise of Human Nature. Ed. L. A. Selby-Bigge, P. Nidditch (2nd ed.).
Oxford; The Clarendon Press.
Kant, Immanuel (Akademie), Kants gesammelte Schriften. Hrsg. Kniglich preussischen
Akademie der Wissenschaften. Berlin: Reimer/de Gruyter, 1902 ff.
- (1997 [1783]; Prol.), Prolegomena to Any Future Metaphysics that may be Able to Come
Forward as a Science. Trans. Gary Hatfield. Cambridge: Cambridge University Press.
Cited following the pagination of the Akademie edition.
- (1997 [1785]; Groundwork), Groundwork of the Metaphysics of Morals. Trans. Mary
Gregor. Cambridge: Cambridge University Press. Cited following the pagination of the
Akademie edition.
- (1998 [1781/1787]; A/B), Critique of Pure Reason. Trans. P. Guyer and A. Wood.
Cambridge: Cambridge University Press. Citations are to the pagination of the first
(A=1781) and second (B=1787) editions.
Kemp Smith, Norman (1923), A Commentary to Kants Critique of Pure Reason. London:
MacMillan.
Kitcher, Patricia (1982) Kant on Self-Identity, Philosophical Review 91: 4172.
- (1984), Kants Real Self. In A. Wood, ed., Self and Nature in Kants Philosophy. Ithaca,
NY: Cornell University Press.
- (1990), Kants Transcendental Psychology. Oxford: Oxford University Press.
- (1995), Revisiting Kants Epistemology: Skepticism, Apriority, and Psychologism,
Nos 29: 285315.
- (1999), Kant on Self-Consciousness, Philosophical Review 108: 34686.
Kitcher, Philip (1975), Kant and the Foundations of Mathematics, Philosophical Review 84:
2350.

Blackwell Publishers Ltd. 2001

Kant, Synthesis, and Cognitive Normativity

305

Kosman, Aryeh (1986), The Naive Narrator: Meditation in Descartes Meditations. In


Rorty, ed. 1986b.
Lange, F. A. (1876), Geschichte des Materialismus und Kritik seiner Bedeutung in der Gegenwart,
3rd ed. Iserlohn: J. Baedecker. Trans. E. C. Thomas. New York: Harcourt, Brace, and Co.,
1925. (N.B., Quoted translation is mine.)
Locke, John (1975), An Essay Concerning Human Understanding. Ed. P. H. Nidditch. Oxford:
Oxford University Press.
Longuenesse, Beatrice (1998), Kant and the Capacity to Judge: Sensibility and Discursivity in the
Transcendental Analytic of the Critique of Pure Reason. Princeton, NJ: Princeton
University Press.
Pippin, Robert (1982), Kants Theory of Form. New Haven, CT: Yale University Press.
Rorty, Amlie Oskenberg (1986a), The Structure of Descartes Meditations. In Rorty 1986b.
-, ed. (1986b) Essays on Descartes Meditations. Berkeley, CA: University of California
Press.
Rorty, Richard (1979), Philosophy and the Mirror of Nature. Princeton, NJ: Princeton
University Press.
Rosenberg, Jay F. (1997), Kantian Schemata and the Unity of Perception. In A. Burri, ed.,
Sprache und Denken/Language and Thought. Berlin: W. de Gruyter, pp. 17590.
Shabel, Lisa (1997), Mathematics in Kants Critical Philosophy: Reflections on
Mathematical Practice. Ph.D. Diss., University of Pennsylvania. Ann Arbor, MI:
University Microfilms.
- (1998), Kant on the Symbolic Construction of Mathematical Concepts, Studies in
History and Philosophy of Science 29: 589621.
Strawson, P. F. (1966, The Bounds of Sense: an Essay on Kants Critique of Pure Reason. London:
Methuen and Co.
Waxman, Wayne (1991), Kants Model of the Mind. Oxford: Oxford University Press.
Windelband, Wilhelm (1884), Kritische oder genetische Methode? In Prludien: Aufstze
und Reden zur Einleitung in die Philosophie, 1st ed. Freiburg i. B. und Tbingen: J.C. B.
Mohr.
Wolff, Robert Paul (1963), Kants Theory of Mental Activity: a Commentary on the
Transcendental Analytic of the Critique of Pure Reason. Cambridge, MA: Harvard
University Press.

Blackwell Publishers Ltd. 2001