THE EPISTEMOLOGICAL UNDERPINNINGS
OF KANT'S SYSTEM
I have therefore found it necessary to deny knowledge, in order to make room for faith. [Kt1:xxx]
Knowledge and Experience
There can be no doubt that all our knowledge begins with experience.... But ... it does not follow that it all arises out of experience. [Kt1:1]
1. The Fundamental Epistemological Distinction
Kant's Critical philosophy is notorious for its terminological ambiguity and apparent inconsistency. The interpretive confusion that often results is at least a contributing factor to the conclusion of many commentators, such as Strawson, that large chunks of Kant's System (e.g., his 'transcendental idealism') are 'unintelligible' and 'incoherent'. Yet I believe, with Kant [Kt1: Axxi], that if his works are approached with 'the patience and impartiality of a judge' (and perhaps even with 'the benevolent assistance of a fellow-worker'), rather than with a set of analytical tools with which to dissect his every sentence, then almost all of his theories can be understood in surprisingly simple and consistent terms. Accordingly, I shall make a further step in this chapter towards the substantiation of this supposition by interpreting and interrelating some of the fundamental epistemological distinctions which serve to structure all three Critiques.
The primary epistemological distinction, underlying directly or indirectly all others in Kant's System, is that between 'knowledge' (Erkenntnis) and 'experience' (Erfahrung). Unfortunately, in spite of (or perhaps, because of) its ubiquity in Kant's writings, it tends to remain an obscure and uncriticized presupposition for both Kant and his many interpreters and critics. The main reason for this neglect seems to be that he invokes a variety of distinctions which define knowledge and experience more precisely, with the result that these more common terms naturally appear to be less technical and in no need of special treatment. The purpose of this chapter will be to explicate the knowledge-experience distinction implicit in Kant's System by integrating it with the most important of these more obviously technical distinctions: first with his pure-empirical and subject-object distinctions, secondly with his a priori-a posteriori and analytic-synthetic distinctions, and thirdly with his distinctions between the empirical, transcendental, logical, and hypothetical perspectives.
The wide range of connotations the terms 'knowledge' and 'experience' have in ordinary language might induce an interpreter to regard any secondary distinctions with suspicion. One might insist that they must inevitably share the indistinct nature of the primary distinction from which they are derived, notwithstanding any intelligibility they seem to have on their own. Such a proposal, however, is unsound; for, as I shall attempt to demonstrate, Kant's own explanations of these distinctions can be interpreted in a relatively clear and plausible fashion. Moreover, even though he does not say much about the terms 'knowledge' and 'experience' as such, he does say enough to provide a basis for a sufficiently coherent interpretation.
In ordinary use, 'experience' can refer generally to the concrete, immediate (i.e., uninterpreted) encounter between a subject and an object. This 'immediate experience' provides the raw material upon which more abstract functions such as 'determinant judgment' and 'reflective judgment' operate [see e.g., Kt7:385-6; s.a. IV.3], for in such experience the subject has neither determined the given object to be an object of knowledge nor reflected upon its epistemological status. Kant uses the word 'immediate' to define 'experience' four times in Kt1, most notably in the Refutation of Idealism, the main purpose of which is to prove that 'outer experience is really immediate ...' But in many other passages he uses 'immediate(ly)' to qualify more specific words such as 'sensation' or 'perception'. In Kt1:A371, for instance, 'immediate' describes both the 'self-consciousness' of 'inner sense' and the 'perception' of 'outer sense'. Moreover, Kant frequently uses the phrase 'possible experience' to refer to this immediate starting point for all knowledge; as such, it denotes 'not only the totality of the objects of experience but also Erfarhen itself as spiritual act' [V4:96; see VII.2.A]. Thus as we shall see, this ordinary, indeterminate and nonreflective sense of 'experience' crops up on numerous occasions throughout Kant's Critical works.
Certainly the most important example of Kant's use of this sense of 'experience' is when he begins the Introduction to Kt1 with the proclamation: 'There can be no doubt that all our knowledge begins with experience' . That is, all knowledge must be related to someone's immediate experience in order for it to qualify as actual knowledge [cf. 629]. Kant stresses the philosophical significance of this type of experience in Kt18:347(82) when he says an 'impression of the senses precedes all the judgments of reason, and carries with it immediate evidence, far excelling all other persuasion.' Yet the priority of immediate experience does not preclude the possibility that certain aspects of our knowledge might be derived from some other source. For several sentences later in Kt1:1 he adds that, although 'all our knowledge begins with experience, it does not follow that it all arises out of experience.'
In explaining how some knowledge is grounded in a source other than immediate experience, Kant does indeed develop another, less conventional, meaning for 'experience'. He describes this more determinate type of experience as a concrete 'synthesis of perceptions' [Kt1:792,161] in which various 'objects of possible experience'  are made actual objects of knowledge through the cooperation of the subject's two main powers of cognitive judgment: intuitive sensibility (which produces sensation) and conceptual understanding (which produces thought). This process, known also as 'determinant judgment', implies a differentiation between two kinds of knowledge: the validity of 'empirical' knowledge is determinable only by appealing at some point to sensible experience [Kt1:2-3; cf. 34], while that of 'pure' knowledge is determinable without reference to sensibility, to the extent that 'there is nothing that belongs to sensation' in it. Kant claims empirical knowledge is tied so closely to experience that the two can, for us, be equated: 'Empirical knowledge is experience.' In this new sense (developed fully only in the second edition of Kt1), experience is no longer the immediate chronological starting point of all knowledge, but one of several 'species of knowledge' [xvii,196; cf. 314]. Unlike empirical knowledge, pure knowledge is related only indirectly to experience: it arises out of the subject's abstract reflection on the general nature of experience [316-9], and is pure in virtue of its primary dependence on the subject rather than the object of knowledge. But in order to engage in such reflection, we must be consciously aware of our experience, not in its immediate state, but as empirical knowledge; for experience in itself is 'the ultimate unconditional given, within which all reflection arises' [G1:348-9].
So far, Kant's use of the words 'knowledge' and 'experience' seems to be relatively clear. The latter refers either to the original encounter between subject and object (i.e., 'immediate experience') which yields actual knowledge through determinant judgment, or to the 'empirical knowledge' which is so produced; and the former refers either to this same empirical knowledge, or to the knowledge which can be inferred from experience by reflecting in other, more abstract ways. But this account of his primary distinction will be of use to the interpreter only if it provides an adequate context for interpreting Kant's secondary distinctions. In IV.2, therefore, I will introduce the four classes of knowledge which arise out of two of his secondary distinctions, after which I will examine in IV.3 the type of reflective perspective which produces each class of knowledge. Finally, in IV.4 I will integrate the various results of this inquiry into a single picture, delineating the essential perspectival pattern which determines the epistemological form of each of Kant's Critical systems.
2. Two Secondary Epistemological Distinctions
The knowledge-experience distinction is rarely discussed as such by either Kant or his commentators because, as mentioned above, experience (even though it has chronological priority in its immediate form) is defined in terms of knowledge. Despite the negligible attention it has been given, this distinction will turn out to form the context in which all Kant's other epistemological distinctions are set. But before this can be fully demonstrated, a good deal more will have to be said about the 'knowledge' side of the distinction. In this section, therefore, I will specify how four basic types of knowledge arise out of the two most prevalent of Kant's secondary epistemological distinctions, between the a priori and the a posteriori and between the analytic and the synthetic, both of which are concerned not only with knowledge, but with the various ways reflective knowledge and immediate experience are related.
On the surface, the bifurcation of knowledge into a priori and a posteriori types seems to be readily comprehensible. A posteriori knowledge is knowledge derived directly from-or the truth of which is contingent upon-the meeting of subject and object in immediate experience. A priori knowledge, on the other hand, is 'given' or 'innate' knowledge which is derived from a source-or the truth of which is-'absolutely independent of all experience' [Kt1:2-3]; hence it is both necessary and universal [3-4]. But upon closer investigation, two problems arise: first, How does this distinction differ from that between pure and empirical knowledge? and secondly, If 'all our knowledge begins with experience', then what sense is there in saying a priori knowledge is somehow 'independent of all experience'? I will address these problems in the following two paragraphs.
Although Kant ordinarily uses the terms 'pure' and 'a priori', as well as the terms 'empirical' (i.e., impure) and 'a posteriori', interchangeably, they should not be regarded as mere synonyms [as in P14:4n], for he does occasionally make a technical distinction between them. The pure-impure distinction discriminates between knowledge which does (impure) and does not (pure) depend directly on some particular sensation, whereas the a priori-a posteriori distinction discriminates between knowledge which is grounded in the subject's experience of some particular object (a posteriori) and that which the subject brings to experience, which must therefore be grounded primarily in the subject itself (a priori). Presumably, knowledge can be a priori even if it involves sensation in some way [Kt1:3,28-9]; or it can be a posteriori without having anything to do with sensation. These possibilities cannot be rejected merely by virtue of the meanings of the words involved. Hence, although examples of pure a posteriori and impure a priori knowledge might be hard to come by, the distinct classes must be acknowledged as logically possible. Nevertheless, they are of minimal importance, since the two pairs are almost always treated coextensively: impure a posteriori knowledge is knowledge derived from a subject's experience of an object (a posteriori) and requiring sensation (impure), while pure a priori knowledge is knowledge brought to experience by the subject (a priori) and requiring no sensation (pure) [x,4-5,124-5].
The status of a priori knowledge in relation to experience should become more evident when I relate the distinctions of this section to the four fundamental reflective perspectives in IV.3. But for now several remarks can be added to dispel some of the ambiguity shrouding the meaning of the word 'knowledge' in the phrase 'a priori knowledge'. Knowledge which arises a posteriori seems not to be troublesome because it is by definition based on experience. A priori knowledge, by contrast, 'which I must presuppose as being in me prior to objects being given to me' [Kt1:xvii; s.a. xviii,xxiii, A128-9], and which is therefore objectively valid 'antecedently to all experience' , is rather more ambiguously called 'knowledge'. This ambiguity can be cleared up by recalling the distinction made in IV.1 between immediate experience (which can lead to 'empirical knowledge') and reflective knowledge (which is known only if experienced, but which might be traceable to some other source). When this is stressed, both a posteriori and a priori knowledge can be regarded as abstractions from immediate experience-though, as will become evident in IV.3, they abstract in different directions. 'A priori' does not denote knowledge which is actually known apart from experience; rather, it refers to knowledge whose validity does not depend on a subject's encounter with particular objects in experience. (Thus, for example, although we cannot actually know, in the a posteriori sense, space and time as wholes, we can claim to have a priori 'knowledge' of these wholes whenever we have a posteriori knowledge of anything.) Kant could have made his meaning less confusing either by not calling the determinate form of experience 'empirical knowledge', or by not using the word 'knowledge' for that which arises out of our reflection on experience. (For instance, he could have referred only to the a priori 'source' or 'conditions' of our actual knowledge.) Using 'knowledge' for both empirical knowledge and its conditions gives rise to uncertainty on the part of the reader as to which sense of the word he intends when he uses it without a qualifier (as in Kt1:26, where he seems to fluctuate between both meanings). Fortunately, once the choices are explicated, the context usually makes his intention sufficiently clear.
Kant also has a more general use for the a priori-a posteriori distinction which should be mentioned briefly at this point. Sometimes he equates all philosophical or 'metaphysical' knowledge with the a priori and all ordinary or 'physical' knowledge with the a posteriori [e.g., Kt7:174,475]. Thus he says 'knowledge through reason and a priori knowledge are the same thing' [Kt4:12; s.a. Kt7:167-8]. This is the sense England has in mind when he says a priori truths 'enable us to understand the "why" of a thing or event', while a posteriori truths 'enable us to know its existence as a fact' [E3:45]. Wolff rightly criticizes Kant's tendency to identify 'the formal (space, time, categories) with the a priori and the material (sensation, empirical concepts) with the a posteriori'; but he goes too far when he adds that this causes Kant to 'be irresistibly drawn to assimilate all knowledge to a priori knowledge.' For Kant's general usage of these terms is never more than a tendency: as we shall see, he is ordinarily very careful to limit the a priori to certain specific sorts of philosophical knowledge, and to allow a definite place for the a posteriori (or material) in his System. Moreover, the dual meaning of 'a priori' (as a specific kind of philosophical knowledge or as all philosophical knowledge) is related to the distinction between Kant's broad and narrow uses of the term 'transcendental' [see e.g., Kt1:80-1], which, as we saw in II.4 and III.4, is quite legitimate once the broad use is understood in connection with the overall Perspective of his Critical philosophy. Nevertheless, since an equivocal use of technical terms is likely to lead to misunderstanding, it should be avoided wherever possible. Accordingly, I shall henceforth treat 'a priori' (and 'transcendental', whenever it is not capitalized) in the narrow sense, and attempt to differentiate more precisely the various sorts of knowledge with which Kant is concerned.
The other important secondary distinction Kant makes between types of (reflective) knowledge is that between 'analytic' and 'synthetic' judgments. Unfortunately, he describes this contrast in a wide variety of ways, which are difficult if not impossible to integrate into a single, consistent picture. Garver, for instance, finds no less than 'twelve theories of analyticity contained in or suggested by Kant's discussion' [G3:245; s.a. V1:1.253f]! Moreover, perhaps as a result of such variety, the nature and validity of this distinction has been a matter of considerable debate in recent years. Although it is not necessary for our purposes to embark on a thoroughgoing study of this topic, examining a selection of the most significant comments of both Kant and his critics will help differentiate Kant's version of the distinction from some of the unkantian versions which have recently been suggested.
Probably the best known of Kant's descriptions of these terms is that in an analytic judgment the predicate is already 'contained in' the subject, while in a synthetic judgment the predicate 'lies outside' the subject [Kt1:10; cf. Kt22:232; see e.g., H4:lii]. A more illuminating, yet less frequently cited description of this distinction is Kant's claim that judgments are analytic only if their truth is 'based entirely on the principle of [non]contradiction', while judgments are synthetic only 'under the condition that an intuition underlies the concept of their subject'. As Allison says: 'Synthetic judgments assert [real] relations [of concepts to objects], while analytic judgments merely assert logical relations between concepts' [A7:54]. With these descriptions in mind, we can use Kant's own pictorial representation of 'particular judgments' in Kt10:103(108-9) (according to which the subject is depicted as a square and the predicate as a circle), to show how (e.g.) 'Yellow is a color' and 'This table is yellow' are propositions representing analytic and synthetic judgments, respectively:
(a) 'Yellow is a color.' (b) 'This table is yellow.'
Figure IV.1: Analytic and Synthetic Judgments
Beck translates Kant's distinction into less metaphorical terms: if '"X is A" implies logically "X is B", the judgment is analytic', but if B is 'related to A by virtue of the fact that both are predicates of the same X', then it is synthetic. Synthetic judgments, then, are informative: they extend our knowledge by providing information about the subject which is not necessarily implied by the meanings of the words (e.g., this table would still be a table whatever its color). Analytic judgments, on the other hand, are, strictly speaking, not informative: the predicate does not extend our knowledge, but provides only what can be inferred from the subject by means of the laws of logic. Although this description of Kant's analytic-synthetic distinction is given predominantly in terms of simple, subject-predicate propositions, it is unfair to charge Kant with limiting his logic to such propositions [as in F6:88]. On the contrary, says Wolff, 'nothing could be further from the truth' [W21:188]. The great variety of applications Kant offers for his analytic-synthetic distinction [s.e. H4:xxii-cxv] is evidence enough of his awareness of the potential complexities of propositional logic. More complex judgments, as Allison explains, are viewed by Kant 'as logical compounds of [these simple] categorical judgments' [A11:25]. Subject-predicate examples, therefore, provide a manageable way of grasping the general characteristics of the analytic-synthetic distinction.
Kant leaves no doubt as to how all this applies to empirical knowledge: 'Judgments of experience, as such, are one and all synthetic' [Kt1:11]. Only when we attempt to interpret such determinate judgments by reflecting upon them does some knowledge come to be regarded as analytic. The bulk of the discussion of the analytic-synthetic distinction by recent philosophers has suffered needlessly by neglecting the implications of this salient qualification. The result has been a running debate over whether the terms refer to a difference of kind or merely to one of degree. The position Kant would adopt on this point becomes evident once his admittedly subtle distinction between immediate experience and reflective knowledge is sharpened (as I am attempting to do in this chapter): both views would be accorded a measure of validity. Kant himself uses the analytic-synthetic distinction primarily as a tool for organizing various forms of reflective knowledge according to their logical status. Thus, his distinction is clearly one between different kinds of knowledge. But, in order to locate the sources of both analytic and synthetic knowledge in immediate experience, this rigid distinction of kind would have to be reinterpreted in terms of varying degrees. The point of Kant's assertion that all judgments of experience are synthetic is simply to emphasize that the term 'analytic' will apply only to certain forms of reflective knowledge, and never to nonreflective experience. It in no way disallows the legitimate formulation of a less restrictive analytic-synthetic distinction, such as those discussed in Appendix IV, in which the terms are not so mutually exclusive.
The main question raised by Kant's introduction of the distinction between analytic and synthetic kinds of reflective knowledge is: How does he intend to integrate it with his distinction between a priori and a posteriori kinds of reflective knowledge? Some philosophers tend to equate the two distinctions (as well as that between 'pure' and 'empirical'); but such an oversimplified approach is not only inadequate [B19:227f], but obviously unkantian [see K2:11.38(Z1:141)]. If, then, the two distinctions are not equivalent, four possible classes of reflective knowledge arise out of their combination: knowledge by reflection might be classified as 'analytic a posteriori', 'analytic a priori', 'synthetic a priori' or 'synthetic a posteriori'. I shall conclude this section by examining briefly what each of these four classes would entail.
To begin with, the impossibility of analytic a posteriori knowledge is generally considered to be 'quite evident' [P5:182-3]: indeed, it is a nonsensical contradiction in terms for those who equate 'analytic' and 'a priori' [see Ap. IV]. Even though Kant argues against those who identify analyticity and apriority [e.g., in Kt1:1-10], he joins them in dismissing this class of knowledge with only a brief explanation: 'it would be absurd to found an analytic judgment on experience. Since, in forming the judgment, I must not go outside my concept, there is no need to appeal to the testimony of experience in its support' [Kt1:11; cf. Kt2:268 and Kt4:12]. There are, however, a few theorists who do regard the analytic a posteriori as providing the best description of certain types of knowledge. Notwithstanding Kant's lack of concern for this class of knowledge, I shall argue in IV.3 that certain aspects of his philosophy can best be understood by reinterpreting them in terms of the analytic a posteriori. At this point, though, it will suffice to say that we should expect such knowledge, if it is possible, to have its validity grounded in some way in experience (a posteriori), and yet also to proceed by making inferences solely on the (analytic) basis of an application of the laws of logic to the concepts or propositions involved.
The second class of reflective knowledge, the analytic a priori, is rather more clearly delineated by Kant. It includes any judgment which, given some previously understood meaning for the terms involved, can be reduced to a logical tautology. This definition of analytic knowledge is what Kant has in mind when he makes such statements as: 'if the judgment is analytic ... its truth can always be adequately known in accordance with the principle of [non]contradiction' [Kt1:190; cf. Kt2:267]. The truth value of such knowledge is both independent of experience (a priori) and determinable solely by the application of logical laws to the concepts involved (analytic).
The third class of reflective knowledge, the synthetic a priori, is by far the most important for Kant, at least within his Critical philosophy [see note III.26]. Indeed, he says in Kt1:19 that 'the proper problem of pure reason is contained in the question: How are a priori synthetic judgments possible?' It is a problem because it is not immediately evident how a judgment could be a priori without being analytic [see Ap. IV]. Such knowledge would be valid independently of any particular experience (a priori) [Kt1:171-2], yet it would also supply new information about the concepts involved (synthetic)-information not deducible by means of formal logic [cf. S15:10 and K9:22-7]. This class of knowledge is of utmost importance to the philosopher because the propositions composing it would be necessarily true without being in any sense trivial: just as the analytic laws of logic determine the general form of what a person can think coherently, so also these synthetic a priori judgments would determine the general form of what a person can experience coherently (i.e., of how a person can convert immediate experience into empirical knowledge). They would therefore provide a solid foundation upon which not only empirical knowledge, but also a philosophical system of knowledge, could rest.
Finally, synthetic a posteriori knowledge is the least troublesome (but also, for Kant, the least philosophically interesting) of the four classes of knowledge. All the knowledge arising out of such empirical factors as scientific experimentation, psychological introspection, the citing of examples, and appeals to 'common sense', falls into this class. Consequently, the word 'knowledge' is usually intended in this sense when it is uttered in ordinary language. Such knowledge consists, quite simply, in judgments which have their validity grounded in various facts of experience (a posteriori), and in which intuitive content is supplied to the concepts involved-content which is not logically implied by the conventional meanings of the words used (synthetic).
3. The Four Reflective Perspectives
The foregoing discussion of Kant's two secondary distinctions between types of knowledge and of the four classes to which they give rise has relied heavily on the supposition that these divisions are intended by Kant as classifications only of knowledge by reflection, and not of immediate experience. In this section I propose to support and enlarge upon this claim by discussing the four methods of reflection, or perspectives, which Kant says can be adopted in considering various objects of knowledge. As suggested in II.4, these will include the empirical, transcendental, logical and hypothetical perspectives, respectively. But first it will be helpful to make some general comments about Kant's use of the word 'reflection'.
Kant distinguishes 'reflection' (Reflexion) from 'comparison' and 'abstraction' by defining it as the act of 'going back over [Überlegung] different representations' in order to determine 'how they can be comprehended in one consciousness' [Kt10:94(100)]. These three 'acts of the understanding' are similar inasmuch as they are all 'logical acts ... by which concepts are generated as to their form' [94(100)]. But elsewhere he puts special emphasis on reflection as the only act by means of which truly philosophical concepts can be generated, for 'reflective judgement' is 'our critical faculty' [Kt7:408; s.a. 395; Kt7i:211; V2:446,451n]. The description 'going back over ...' implies that the representations which give rise to various philosophical perspectives have already been 'gone over' once. This indeed is precisely what Kant intends to get across by his distinction between 'determinant [bestimmende] judgement' and 'reflective [reflectirende] judgement' [Kt7:385-6]. 'Determinant judgment', interprets van de Pitte, 'is constitutive of the world of factual experience and is thus objectively valid. Reflective judgment, on the other hand, is merely an interpretive technique which we employ in order to bring organic entities and systematic unities within our powers of comprehension. It thus carries only a subjective validity.' This distinction is closely related to that between immediate experience and reflective knowledge: determinant judgment converts immediate experience into empirical knowledge by subsuming a particular intuition under a universal concept, and reflective judgment converts empirical knowledge into more abstract forms of reflective knowledge by positing the universal which serves as the guiding principle for a given set of particulars [Kt7:179-80]. With this distinction clearly in mind, we can now examine the nature of the four fundamental perspectives which operate throughout Kant's System.
In the first two Critiques Kant does not use the word 'reflection' as a technical term for the activity of viewing objects from an empirical perspective. Instead, he uses phrases such as 'the empirical employment of understanding' or 'the empirical employment of reason' [see II.3.C] whenever he wishes to describe some aspect of the empirical perspective as it operates in one of these systems. (Many of the empirical elements introduced in these systems are presented merely as by-products of other perspectives [e.g., Kt1: 152; Kt5:390]; but the constitutive role of the empirical perspective in systemt will be discussed in VI.3 and VII.3.A, and that of systemp, in VIII.3.A.) However, Kant sometimes mentions in passing the role of reflection in the empirical perspective, as when he describes an 'empirical deduction' as one which 'shows the manner in which a concept is acquired through experience and through reflection upon experience' [Kt1:117; s.a. 503].
In the third Critique, by contrast, Kant's use of the phrase 'reflective judgment' is, as Evans argues, equivalent to his former use of the phrase 'empirical employment of pure reason' [E4:483; s.a. G6:457], thus implying that the perspective which determines Kant's standpoint for examining such judgment in systemj is the empirical. As we saw in II.4, each of Kant's three systems adopts one perspective in this way as the standpoint to guide the operation of all four perspectives in that system. Although my main focus in this section will be on their role as perspectives in systemt, I will also refer at several points to the way in which each forms the basis for a discrete standpoint. This will prove to be especially important in our discussion of Kant's hypothetical perspective. (In Part Three we will discuss various ways in which the four perspectives change when they are applied from standpoints other than the theoretical.)
A person who adopts an empirical perspective reflects upon particular objects of experience without attempting to 'go beyond' their nature as given in immediate experience. In empirical reflection as such there is no need to discriminate between the respective roles of the knowing subject and the known object, because the two are fused in experience. This continuity between immediate experience and knowledge resulting from empirical reflection is, no doubt, what leads Kant to make the (potentially misleading) claim that 'empirical knowledge is experience' [see IV.1]. Strictly speaking, 'empirical knowledge' should denote only that synthetic a posteriori knowledge which arises out of empirical reflection on the objects of one's experience. Thus, empirical knowledge of 'cause', for instance, refers neither to the actual (i.e., immediate) experience of some particular cause, nor to the ability to determine its subjective or objective ground; rather it consists in the ability to answer the question 'What is the cause of X?' by thinking and reasoning straightforwardly about the objects of one's experience.
Ordinarily, we do not distinguish between our experience and our reflection on experience, since any type of reflection must itself be part of our immediate experience in order to bring forth knowledge which is actually known [see IV.2]. Thus, in everyday life all reflective perspectives tend to be mixed indiscriminately. (This is the situation, incidentally, which gives rise to the need for a Transcendental Perspective as the foundation for a philosophical System, within which our various perspectives can be distinguished in an orderly fashion.) All types of reflective experience attempt to give elegance to their vulgar counterpart, nonreflective experience. In the case of empirical reflection, the transition from vulgarity to elegance tends to be gradual, because of the affinity between immediate experience and empirical knowledge-i.e., because of the need to appeal to our sensible experience whenever we try to establish synthetic a posteriori knowledge. But in each of the other three types of reflection, to which we shall now turn our attention, the qualitative transition tends to be rather more abrupt.
Of all the perspectives in Kant's System, the transcendental perspective plays the most important role [Kt1:25-6; cf. P2:1.226-30 and E5:29]. Indeed, the a priori-a posteriori distinction itself first arises in this context. Unfortunately, the fundamental significance of the 'transcendental reflection' with which this new perspective is concerned could be easily overlooked by the reader, because Kant waits until an Appendix in the middle of Kt1 to discuss its importance in detail. The reason he waits until this point is that, before he can show how transcendental reflection reveals the errors in any non-Critical philosophy [see note IV.25], he first has to specify the doctrines which can be established by adopting this transcendental alternative. But this gives the misleading impression that transcendental reflection is more a convenient tool for the comparison of various treatments of specific philosophical issues than (as we have seen in II.4, III.4 and IV.2) the essential methodological tool defining the overall Perspective for all three Critical systems!
Kant does, however, give one of his clearest accounts of what the transcendental perspective entails as early as Kt1:25 [cf. 185,196-7]: 'I entitle transcendental all knowledge which is occupied not so much with objects as with our Perspective on knowledge of objects in so far as this Perspective on knowledge is to be possible a priori. A system of such concepts might be entitled transcendental philosophy.' Kant elsewhere says his task as a transcendental philosopher is to 'enquire what are the a priori conditions upon which the possibility of experience rests, and which remain as its underlying ground when everything empirical is abstracted from appearances [i.e., from the objects of experience].' A transcendental perspective, then, presupposes the subject-object distinction: it attempts to determine what there is in the subject a priori which makes possible our knowledge of the objects we experience. Because these conditions must be added by the subject to the objects given in intuition to produce such empirical knowledge, they are (both logically and methodologically) synthetic as well as being a priori. That the knowledge arising out of this radically epistemological perspective concerns only a set of synthetic a priori forms embedded in the subject is spelled out explicitly by Kant when he says 'the word "transcendental" ... never means a reference of our knowledge to things, but only to the cognitive faculty' [Kt2:293; s.a. Kt1:74-5].
When Kant finally gets around to describing what transcendental reflection is, he says it is the act of determining 'in which faculty of knowledge [given representations] belong together subjectively-in the sensibility or in the understanding' [Kt1:317]; in so doing one determines whether or not each representation is pure. Accordingly, such reflection is the necessary first step in adopting a transcendental perspective; for it would be impossible to abstract everything empirical from experience without first differentiating between what is pure and what is impure (i.e., empirical) [80-1]. But in a broader sense [see II.4 and IV.2], all the steps involved in determining the synthetic a priori forms of empirical knowledge can be regarded as arising out of transcendental reflection. Thus, transcendental knowledge of 'cause', for instance, refers neither to the actual experience of some particular cause, nor to the ability to determine such a cause through empirical reflection; rather, it consists in the ability to answer the question 'What is the status of causality in the general relation of a subject to an object?' by reflecting transcendentally on the synthetic a priori conditions for the possibility of experience.
Two remaining points should be made concerning the transcendental perspective to help guard against possible misunderstanding. First, some common uses of 'transcendental', according to which the word refers to a special kind of consciousness, or to 'the grasping of things as they are in themselves' [M12:163], or even to 'God's point of view' [C9:84], might lead to the mistake of confusing the transcendental perspective with the 'ivory tower' perspective of typical non-Critical metaphysicians, who assume they can ascend reflectively to such heights as to attain a perfectly objective view of transcendent reality. Kant, having devoted the bulk of the Dialectic in Kt1 to the task of disclosing the error inevitably bred by this 'logic of illusion' , explicitly rejects this interpretation in Kt2:373n: 'High towers ... are not for me. My place is the fruitful bathos of experience ...' Indeed, such error is precisely what he believes he can avoid by emphasizing the differences in the various perspectives which can be adopted legitimately in the quest for knowledge. By referring to the synthetic a priori as 'knowledge', he is not claiming to possess a special type of knowledge which is actually known independently of the limitations of experience; rather, like all knowledge, it can be known only when a person experiences a certain kind of reflection. Kant supports this point when, in response to a misunderstanding of his use of the word 'transcendental', he says it 'does not signify something passing beyond all experience but something that indeed precedes it a priori, but that is intended simply to make knowledge of experience possible' [Kt2:373n]. When properly understood, adopting the transcendental perspective can be seen not only to be legitimate, but to be the 'duty' of the philosopher. Far from being a kind of 'ivory tower' perspective, it determines the epistemological foundations on which our knowledge and experience is built [Kt1:195], and in so doing, reveals that all human knowledge is inextricably tied to certain limits it cannot transcend.
The second point is that Kant does not limit synthetic a priori knowledge to the philosopher. On the contrary, as suggested in IV.2, there is a sense in which anyone who has any empirical knowledge must also (unconsciously) possess transcendental 'knowledge'. For instance, Kant says that 'in all theoretical sciences of reason synthetic a priori judgments are contained as principles' [Kt1:14]. Viewing such principles from a transcendental perspective is important (philosophically) because it is only through transcendental reflection that their status can be shown to be synthetic a priori [cf. 81,316-7,749-50]. The extent to which mathematicians, for example, know their principles to be synthetic a priori is the extent to which they have reflected transcendentally on their status. But the word 'know' here refers only in a loose sense to 'transcendental knowledge', insofar as the latter can refer to the (empirical) knowledge that a given proposition is synthetic a priori.
Distinguishing between the empirical and transcendental perspectives is recognized by many recent commentators as being essential to an adequate understanding of Kant's Critical philosophy [see II.2]. Unfortunately, these commentators usually emphasize this distinction so much that another, equally important perspectival distinction tends to be ignored [see e.g., B20: 36-51,140-8; A6:194]. Although it is true that most of the problems Kant attacks in Kt1 are, as Allison says, solved 'by means of the perspectival conception of the relation between the transcendental and the empirical' [A6:203; but see Kt1:189-91], the distinction between the logical and the hypothetical perspectives is, as I shall demonstrate in the remainder of this section, just as vital to the success of Kant's System. The importance of the latter distinction is often recognized only as it applies to the standpoints in Kant's System, since the logical and hypothetical perspectives give rise, respectively, to the theoretical and the practical standpoints [see II.4]. For instance, Wolff has this distinction between standpoints in mind when he rightly says that for Kant reason is 'the faculty both of logic and of ethical judgment' [W21:204]. As usual, Kant is partly to blame for this interpretive problem, since he often mixes the terms which properly refer to perspectives with those which properly refer to standpoints, as when he says in Kt10: 72(80) that 'all our conviction is either logical or practical.'
Immediately after introducing 'transcendental reflection' as a technical term, Kant contrasts it with 'logical reflection'. He says at this point only that the latter 'is a mere act of comparison' which takes 'no account whatsoever of the faculty of knowledge to which the given representations belong' [Kt1:318]. That is, from a logical perspective, there is no need to determine whether the objects of reflection 'are noumena for the understanding, or are phenomena for sensibility' , because all that matters is their compatibility with the laws of logic [189-91]. The logical perspective is the 'merely formal' employment of reason which 'abstracts from all content of knowledge' . Logical reflection is like all types of reflection, however, in being ultimately dependent on the 'possibility of experience' [195; s.a. Kt15:81(244)]. It is similar to empirical reflection in that it operates without distinguishing between the subject and object of experience; and it is similar to transcendental reflection in that it seeks to establish a priori truths; but it is different from both in that it 'has nothing to do with the origin of knowledge, but only considers representations ... according to the laws which the understanding employs when ... it relates them to one another' [Kt1:80]. This means the aim of logical reflection is always analytic: it is concerned only with determining whether or not the representations in a given proposition are related in a form which can be reduced to a tautology [see Z2:169-70]. The tools used in such reflection are those enumerated by what Kant calls 'pure general logic' [Kt1:78], and the goal towards which it works is the systematic delineation of the analytic a priori knowledge which is applicable to specific sciences .
Just as the a priori-a posteriori distinction makes sense only if one engages in transcendental reflection, the analytic-synthetic distinction makes sense only if one engages in logical reflection (yet, once made, both distinctions relate to the classes of knowledge which arise in all four reflective perspectives); for as Schulze accurately declares, the latter 'division is itself derived immediately from the principle of [non]contradiction' [S7:174; cf. P8: 98-9]. Kant is careful to point out that a proper understanding of the implications of this distinction requires transcendental reflection as well, since general logic is unconcerned with the synthetic a priori [Kt1:824; Kt22:242-5]. But this in no way detracts from the need to stress the logical character of its analytic side in order to bring out the difference between it and the empirical versions of the distinction [see Ap. IV]. For Kant, the status of a proposition can be determined to be analytic only through logical reflection [see O1:336]; therefore, a proposition considered (by means of transcendental reflection) to depend on some synthetic element, such as intuition, may or may not be logically analytic. Consider, for example, the question 'How do you know all bachelors are unmarried?' We cannot show our knowledge to be logically analytic by appealing to experience and answering 'Well, all the bachelors I've ever known, now that I think about it, have been unmarried, therefore ...' ( la Quine), or even by answering 'Being a bachelor is always connected by linguistic convention with being unmarried, therefore ...' ( la Bird); the only way to show such knowledge to be logically analytic would be to answer 'If (given a previously agreed upon use of terms) we map that proposition onto the laws of logic, it eventually reduces to a tautology, therefore ...' ( la Kant). Logically analytic truths might be employed in the context of an empirical version of the analytic-synthetic distinction ( la Bennett [see Ap. IV]), but no one could know they are analytic without engaging in logical reflection.
The fourth and final type of reflection is that which aims to produce 'practical knowledge' when a subject adopts the hypothetical perspective. The best way to back up my proposal that hypothetical/practical reflection should be regarded as the correlate of logical reflection in a way comparable to the transcendental-empirical correlation would be to show from Kant's own words that it yields the one class of knowledge which has so far gone unmentioned in this section, the analytic a posteriori. At first sight, this alternative seems to be precluded by his hasty rejection of the possibility of such a class of knowledge [Kt1:11] and by his unfortunately broad understanding of a priori knowledge, according to which it refers not only to the knowledge yielded by transcendental or logical reflection, but to that which is necessary in any non-physical sense [see e.g., 661; cf. IV.2]. The matter is further complicated by the fact that, although he intends his practical standpoint (based as it is on the hypothetical perspective) to replace the traditional form of metaphysical reflection, he never makes it entirely clear just how the status of the knowledge yielded by the former differs from that yielded by the latter. I will therefore examine first Kant's view of the status of traditional metaphysical reflection, secondly, how this differs from his own hypothetical perspective, and thirdly, how the latter gives rise to the practical standpoint. Only then will we be prepared to make an adequate assessment of the extent to which the analytic a posteriori class of knowledge might find a place in Kant's System.
As early as the Introduction to Kt1 Kant states unambiguously that 'metaphysics ... ought to contain a priori synthetic knowledge. For its business is ... to extend our a priori knowledge' [Kt1:18; cf. Kt2:273-4]. He later adds that the metaphysician cannot obtain this goal 'by mere [or 'naked' (bloss)] reflection', but only by clothing it with 'inference' [Kt1:366]. Inference is required because 'the [metaphysical] concepts of reason ... are concerned with something to which all experience is subordinated, but which is never itself an object of experience'-namely, 'the unconditioned' . In itself, the unconditioned is, as Allison points out, 'an analytic principle, depicting what is contained in our concept of a thing in general.' Because it is a pure concept 'transcending the possibility of experience', Kant calls it an 'idea' [377,382-3]. But the metaphysician who attempts to use such ideas synthetically to make inferences without first engaging in transcendental reflection is likely to assume that synthetic a priori judgments can apply directly to the unconditioned, as if it were an intuitable object of ordinary experience [325-6,410,662-3]. The 'misinterpretation' of the 'concepts of reflection' , which characterizes this 'speculative perspective' , inevitably leads to the sort of ambiguity and illusion which Kant attempts to dispel in the Dialectic [354-5]. In each case the fallacy arising out of speculative reflection has the same essential character: metaphysical reflection which has not been limited by a prior use of transcendental reflection will be patterned solely along the lines of a pseudo-transcendental mixture of empirical and logical reflection; that is, it will attempt to produce synthetic a priori knowledge by conflating the logical perspective and its a priori aspect with the empirical perspective and its synthetic aspect.
Within the confines of systemt, Kant does not offer a clear alternative to speculation as a means of establishing positive metaphysical conclusions. Instead, he argues throughout the Dialectic, but especially in the final Appendix [Kt1:670-732], that as long as we limit our attention to the theoretical standpoint (i.e., by assuming the primacy of the logical perspective), the only proper way of dealing with metaphysical questions regarding the three 'transcendental ideas' [383-4]-viz., 'God', 'freedom', and 'immortality'-is to adopt the hypothetical perspective. As he explains in Kt1:675: 'Reason's hypothetical perspective is regulative only', rather than constitutive. This means the hypotheses we form about such ideas can never be proven to be true, but can only be viewed 'as if' they are true [e.g., 698-701]. Because each idea is 'a necessary concept of reason to which no corresponding object can be given in sense-experience' , Kant stresses that in systemt 'it remains a problem to which there is no solution' [384; s.a. 809]. Although he never assigns a particular epistemological status to the hypothetical use of ideas, as he does for speculative knowledge, he does allude to the affinity (and opposition) between the logical and hypothetical perspectives: he warns that this 'problem' is to discover something like 'a logical principle', rather than 'a transcendental principle', of ultimate unity: 'Reason's hypothetical perspective has ... as its aim the systematic unity of the knowledge of understanding' [675-6]. This differs from the logical perspective as such because for the hypothetical perspective such unity is 'set as a task' for reason, whereas for the logical perspective it is 'already given' [526-7].
In the Doctrine of Method of Kt1 Kant devotes an entire section [797-810] to the issue of the nature and function of the hypothetical perspective in systemt, as well as its relation to the logical and transcendental perspectives. He gives two conditions under which an opinion can serve as an hypothesis : (1) 'the possibility of the object' to which it refers 'must ... be ... completely certain'; and (2) the 'opinion ... must be brought into connection with what is actually given ... Then, and only then, can the supposition be entitled an hypothesis.' After warning against the danger of inventing hypotheses about new types of entities which somehow do not conform to the synthetic a priori conditions for the possibility of experience (viz., space, time and the categories [see VII.2]), he then explains that the proper role of hypotheses in systemt is only to aid 'the understanding in the field of experience' by providing 'regulative principles' of systematic unity [798-800]. This means it is 'permissible to think' of the ideas as real entities, but not 'to assume' that they could ever be presented as such in experience. The latter would require a 'transcendenal hypothesis', which 'would really be no explanation' , and which 'can never be permissible from reason's speculative perspective' , whereas the former requires only the quasi-logical recognition that the ideas are at least permissible as 'heuristic fictions'.
Instead of labelling such hypotheses with a specific epistemological status, he says they provide not even 'the least knowledge', since 'we are not actually asserting' anything in forming an hypothesis [Kt1:808]. If we neglect this fact by misusing an hypothesis in a dogmatic fashion (i.e., by assuming the idea is 'necessary, not only subjectively and logically, as method, but objectively also' ), the result is that it 'relieves us from further investigation, and our enquiry is brought to an end [but] not through insight' [801-2]. Nevertheless, Kant goes on to argue that these same hypotheses can and ought to be used 'in a polemical fashion' in order to nullify 'the sophistical arguments by which our opponent professes to invalidate' the reality of the ideas . 'Hypotheses are therefore, in [systemt], permissible only as weapons of war, and only for the purpose of defending a right, not in order to establish it' by 'proof' . By using hypotheses in this 'merely negative' fashion  to show the skeptic that he 'has too little understanding ... to allow of his flattering himself that he has the advantage in respect of speculative insight' , 'we are ... proceeding in entire conformity with reason' . Indeed, 'for our complete equipment we require ... the hypotheses of pure reason' [806e.a.]. The reason, Kant explains, is that this hypothetical perspective serves to protect, from within the theoretical standpoint, the 'interest of reason' which the speculative perspective is powerless to defend by means of proofs [see e.g., 676 and 803]. It must perform this task because, although the hypothetical perspective in systemt can be employed only negatively, 'reason has, in respect of its practical standpoint, the right to postulate what in the field of mere speculation it can have no kind of right to assume' .
In order fully to understand the nature of Kant's hypothetical perspective, therefore, we must examine how he uses it in the form of the practical standpoint to propose a more positive alternative to the traditional type of metaphysical reflection. The speculative assumption that logic and its theoretical standpoint provide the foundation for metaphysical reflection must, he maintains, give way to the new assumption that practical reflection provides the only secure foundation for metaphysics. Just as the logical perspective guides the search in systemt for the elements of pure theoretical reason whose application is necessary for the possibility of natural experience, so also the hypothetical perspective guides the search in systemp for the elements of pure practical reason whose application is necessary for the possibility of moral experience. The former elements compose the Critical foundation for a 'metaphysics of nature' and define Kant's theoretical standpoint, while the latter compose the Critical foundation for a 'metaphysics of morals' and define Kant's practical standpoint [Kt5:388; see III.4].
Although Kant is unclear about the epistemological status of hypothetical reflection, he does describe the status of practical reflection in a way which highlights his claim that it can replace the illusory synthetic a priori knowledge of speculative reflection. In Kt1:xxi he seems to take it for granted that practical reflection ought, like transcendental reflection, to yield synthetic a priori knowledge, but he explains that such 'practical knowledge' is a priori 'only from a practical standpoint' [s.a. 364,691], which leaves open the question of what its status would be when viewed from the theoretical standpoint. Clearly, Kant has to regard the elements in systemp, like all elements in his System, as synthetic a priori in the sense that they are governed by the overall Transcendental Perspective [see II.4]. Yet he does not put nearly as much emphasis on demonstrating this status in systemp as he does in regards to the elements in systemt. His failure to explain how the synthetic a priori status of practical reflection relates to the synthetic a priori status of transcendental (or speculative) reflection runs the risk of obscuring several major differences between these perspectives, which have to be ignored or underemphasized if one is to continue thinking of them as yielding the same class of knowledge. Yet such differences do exists, for why else would Kant have added 'though only from a practical standpoint' in Kt1:xxi?
What would it mean to view 'practical knowledge' from a theoretical standpoint? Since Kant views the practical as reason's proper standpoint and the hypothetical as reason's proper perspective in systemt, he would no doubt begin to answer this question by saying the hypothetical perspective is viewing practical knowledge from the theoretical standpoint. And in Kt1:691 [s.a. 697-8] Kant himself lists three differences between hypothetical principles and ordinary transcendental principles: (1) 'A transcendental deduction cannot ... be effected' in respect to practical ideas; (2) rather than constituting empirical knowledge, they serve merely 'for the guidance of the empirical perspective'; and (3) their 'validity' is 'objective', but 'indeterminate' rather than determinate. Such important differences should have been enough to convince Kant of the need to classify the legitimate use of the hypothetical perspective (i.e., any theoretical view of practical knowledge) with a special status of its own. Instead, as we have seen, he neglects this issue, probably because he knew the hypothetical perspective as such does not produce knowledge. Another reason for this neglect could be that the hypothetical perspective is designed explicitly to point beyond itself to the practical standpoint. For as Kant explains in Kt4:5, the 'problem' of the 'hypothetical' use of ideas in systemt 'now [in systemp] becomes an assertion' which can fully satisfy the 'need' of reason out of which the ideas arise-i.e., the hypothetical perspective itself attains in systemp the status of synthetic a priori knowledge. Nevertheless, the converse of this is that the practical principles in systemp can be viewed from the theoretical standpoint, and that they ought therefore to have some special status when viewed in that way. Surprisingly, here in Kt1:691 Kant actually suggests that, despite the differences, hypothetical principles 'seem to be transcendental', since they consist of 'synthetic a priori propositions'.
Whenever Kant regards the products of hypothetical and practical reflection as 'synthetic' and 'a priori', he must be allowing these terms to take on significantly new meanings. Indeed, 'a priori' is no longer being used to define a necessary and universal subjective condition for the possibility of experience; it is now being used to define an idea, which is a 'concept of reason' inferred by the subject from experience [Kt1:367-8], which plays no part in making that experience (i.e., empirical knowledge) possible, and whose application to immediate experience is neither necessary nor universal. As Kant himself says: 'No actual experience has ever been completely adequate to [the idea], yet to it every actual experience belongs' , for it is 'so constructed as necessarily to contain the concept of what is absolutely primordial' . The claim that all experience is contained in this concept of reason implies not that such an idea is a priori (since the idea is certainly not given before experience) but that it is analytic [cf. A8:237-8]. This seems especially evident when Kant tells us that synthetic knowledge through reason is 'completely impossible', since 'understanding alone is capable of true synthetic a priori items of knowledge' [Kt1:824], and that 'the criterion of the possibility of synthetic knowledge is never to be looked for save in experience, to which the object of an idea cannot belong' [630e.a.]. If the idea cannot produce synthetic knowledge, then it surely ought to be regarded as analytic.
The term 'synthetic' must likewise refer no longer to knowledge whose truth is verified by appealing to some factual, intuitive content [cf. A7:72-3], but to knowledge whose truth is dependent on its compatibility with various theoretical 'hypotheses' [Kt1:797-810] or practical 'laws' [Kt5:388]. Kant explains that 'in the practical standpoint, our sole concern is with the carrying out of rules' [Kt1:384-5e.a.]-a concern for the instantiation of practical ideas in experience which would seem to be more a posteriori than synthetic. This seems especially evident when we consider it together with his claim that 'the indispensable condition of reason's entire practical standpoint' is that 'the idea of practical reason can always be given actually in concreto, although only in part' [384-5]. If the idea cannot be given as an individual whole, which is a characteristic of all intuitions [see VII.2.A] and thus a criterion of syntheticity, then its in concreto character must surely refer to some non-synthetic type of a posteriori givenness.
If I am right in pointing out these shifts in meaning, then, whenever Kant says something like 'X is synthetic a priori, though only from a practical standpoint', we can interpret this as meaning 'X is analytic a posteriori'. For the changes he makes to his ordinary sense of 'a priori' when he applies it to the practical standpoint actually convert it into his strict sense of 'analytic', and those he makes to his ordinary sense of 'synthetic' convert it into his strict sense of 'a posteriori'. Why does Kant neglect these discrepancies in his usage? The explanation I would offer is that, since his ultimate goal is to defend rather than to destroy many of the traditional doctrines of metaphysics [Kt1:xxiv-xxxi; see X.1], he might have thought (in keeping with his rationalist background) that to give a different status to practical knowledge would set it too far apart from that of traditional metaphysical knowledge; so instead, he inadvertently altered the meanings of his terms and (supposedly) preserved the same status.
Now this terminological change may seem drastic, yet I do not believe it entails any substantial revision of the various theories Kant puts forward in the Dialectic of Kt1 or in his moral philosophy, especially since, as we have seen, he himself makes contradictory claims about the status of hypothetical and practical reflection. On the contrary, referring to the knowledge arising out of hypothetical and practical reflection as 'analytic a posteriori' clarifies its status by abiding more strictly than Kant himself does to the meanings he originally assigns his terms. I explore in Appendix IV the nature of this type of knowledge in a bit more detail by relating it to the views of several other philosophers. At this point, however, let us examine more closely the implications of this change for Kant's System.
Just as Kant discusses the close correlation between the hypothetical and logical perspectives in Kt1:676, he compares the practical standpoint to logic in Kt35:(2-3): 'like logic, practical philosophy does not concern itself with a particular sort of cases of practical activity but deals with the practice of free actions in general without reference to any case whatsoever.' Both logical and practical reflection depend on certain laws, either 'of the understanding' or 'of the will'. Just as the highest principle of logical reflection is revealed in systemt to be the law of noncontradiction [Kt1:189; cf. III.3], so also the highest principle of practical reflection is revealed in systemp to be the law of 'duty' [Kt5:397; Kt4:32] or 'the categorical imperative' [Kt5:420; Kt4:21]. In both cases the law is analytic in relation to other laws in its system because it can be used to test the validity of such subordinate laws, yet it cannot itself be verified by appealing to a higher law [see note IV.18]. The difference is that, whereas logical laws are necessary a priori for all thinking and are thereby equally applicable in principle to all experience, even in systemt, practical laws apply to what ought to be the case, a posteriori, in 'matters of conduct' [Kt1:575], and 'allow for conditions under which what should happen often does not' [Kt5:388]. Thus, for example, we call someone 'good' by judging the extent to which their behavior, considered a posteriori, coincides analytically with the idea of 'perfect goodness'-that is, the extent to which their behavior is, as it were, 'contained in' that idea of perfection. Likewise, in systemt the unconditioned ideas of reason must be presupposed to refer to the analytic totality of some empirical synthesis [Kt1:701; Kt4:132,134; see VII.3.B], so our knowledge of them can be described most adequately as the analytic a posteriori counterpart of the analytic a priori knowledge gained through logical reflection.
Interpreting both the hypothetical perspective of systemt and the practical standpoint of systemp in terms of analytic a posteriori knowledge does not imply that the 'ought' of practical reflection is determined by the empirical 'actions and conditions of the human volition'-a view Kant explicitly denies [Kt5:390; cf. Kt1:575-6]. Rather, I am arguing that when, for example, Kant says our awareness of an '"ought" expresses a possible action the ground of which cannot be anything but a mere concept' , the part he thinks makes such knowledge synthetic (i.e., its appeal to a 'possible action') actually makes it a posteriori, and the part he thinks makes it a priori (i.e., the fact that its 'ground' is a concept such as 'God' or 'goodness', rather than an appearance of God or of something good) actually makes it analytic. How else could a standpoint be 'practical' except by stipulating an analytical connection between an abstract concept and an a posteriori experience which ought to be subsumed under it?
If Kant's practical standpoint is to be regarded as yielding knowledge which is analytic and a posteriori, then what sense can be made of his assertion, quoted in IV.2, that 'it would be absurd to found an analytic judgment on experience'? Admittedly, this statement indicates that Kant saw no use for the label 'analytic a posteriori' as describing a class of knowledge. Such an attitude results from his tendency to limit the use of 'analytic' to the knowledge arising out of the logical perspective, and that of 'a posteriori' to the knowledge arising out of the empirical perspective. Had he considered the possibility of describing the knowledge produced by hypothetical and practical reflection as analytic a posteriori, he could have reworded his rather extreme condemnation in such a way as to bring out his two essential points more clearly: first, that it would be absurd to found any logically analytic knowledge on experience (because it is a priori); and second, that the need to found practically analytic knowledge on experience makes it impossible ever to reach the 'absolute certainty' which is possible in some other types of reflection, so that in a sense it is absurd to regard it as 'knowledge'. Is not this latter assertion just what Kant is alluding to when he says he intends 'to deny knowledge, in order to make room for faith' [Kt1:xxx]? Indeed it is. And when this is recognized, the audacious claim that practical reflection is concerned with the analytic a posteriori becomes rather more tame.
Kant's whole point in the Dialectic of Kt1 is to demonstrate how the limitations revealed by transcendental reflection prevent the metaphysician from attaining knowledge through speculative reflection. In place of the latter he proposes the need for a reasoned faith in the (analytical) ideas as subjectively necessary presuppositions of the hypothetical perspective in systemt. An idea on its own always remains a 'problematic concept' [Kt1: 445n]. So we must believe it to be true even though it lies 'out beyond' the limits of the empirical perspective . To do so is to view it 'as if' it were analytically applicable to all experience [698-703]. Likewise, viewing myself in systemp as if I am a free agent (i.e., believing I am such) is the only way I can coherently explain my actions as being 'moral' [see e.g., Kt5:459,462-3]. Kant says such practical presuppositions, whether in systemt or systemp, do not 'extend' my knowledge in any way, for 'no synthetic proposition is made possible by conceding their reality.' But neither does the 'as if' commit me to believe in a mere 'philosophical fiction' [see note IV.35]. On the contrary, it connotes that, although the transcendental limits of my experience in systemt make it impossible for me to have empirical knowledge that I am free, the practical limits of my action in systemp provide very good reasons for adopting a rational belief that I am free: namely, that from the hypothetical perspective I must assume my (a posteriori) experience of morality is contained analytically within the notion of freedom [cf. Kt1:702-3 and Kt4:132,134]. Accordingly, the most accurate statement of Kant's position is that, whereas speculative reflection attempts to establish the synthetic a priori status of metaphysical knowledge-claims, hypothetical/practical reflection admits that the epistemological status of such claims cannot (and need not) be anything other than analytic a posteriori belief.
4. A Summary and Model of Kant's Reflective Method
In the second edition Preface to the first Critique Kant stresses that his book 'is a treatise on the method' through which 'the procedure which has hitherto prevailed in metaphysics' can be revolutionized, 'and not a system of science itself.' If we interpret this claim literally-and I am convinced we should-it means Kant's main goal is not (as is commonly supposed) to establish a particular set of transcendental 'principles'. This is undoubtedly one of his most interesting and influential secondary aims; but, as Vleeschauwer observes, Kant always tends to approach the subject with which he is concerned 'with the clearly avowed intention of showing how ... everything depends on method' [V4:19; cf. W5:97]; so his main purpose must be to delineate the patterns of thinking which the philosopher must adopt in order to construct a coherent philosophical system. We examined the form of such patterns in detail in III.3. Our findings in the foregoing section have now revealed one of Kant's most important methodological applications of the fourfold (2LAR) pattern, according to which it defines the relationship between four methods of inquiry, or reflective perspectives, by means of which we can attain knowledge. In fact, one's ability to understand the significance of the particular arguments or theories in Kant's System of Perspectives is likely to be directly proportional to one's grasp of the implications of this fundamental pattern.
In this chapter I have attempted to uncover this key epistemological pattern by investigating Kant's various ways of distinguishing between 'knowledge' and 'experience'. The results can be summarized as follows. 'Immediate experience' refers to an indeterminate, nonreflective encounter of subject and object in the ordinary world. 'Knowledge' refers to the results aimed at when a person chooses to assume one of four 'perspectives' on that experience by engaging in one of four corresponding types of 'reflection'. To adopt an empirical perspective is to reflect on what 'is true' about one's experience without taking note of the distinction between the subject and object of knowledge; its goal is to reach synthetic a posteriori knowledge. To adopt a transcendental perspective involves distinguishing clearly between the subject and object in order to reflect on the subjective conditions which 'must be true' in order for it to be possible for a subject to experience an object; its goal is to reach synthetic a priori knowledge. To adopt a logical perspective involves abstracting completely from the subject-object distinction in order to reflect on what 'must be true' because the logical laws of thought require it to be so; its goal is to reach analytic a priori knowledge. To adopt a speculative perspective involves distinguishing between the subject and object, but ignoring the role of the subject, in a fallacious attempt to reach synthetic a priori knowledge of the thing in itself, even though such knowledge extends beyond the limits set by transcendental reflection. The hypothetical perspective replaces the speculative perspective by distinguishing properly between the subject and object as in transcendental reflection, and then reflecting on what we can reasonably treat 'as if true' about both transcendent reality and experience in light of the requirements of systematic unity in systemt [see VII.3.B]; Kant vaguely suggests its goal is to reach the synthetic a priori, but I have argued that he really means its goal is to reach analytic a posteriori belief.
Kant's Four Reflective Perspectives on Experience
The most effective way of demonstrating the integrative coherence of this pattern is to plot all these terms and their intricate relationships onto a single, schematic 'map' of Kant's perspectival methodology, as shown in Figure IV.2. The center of this map is occupied by immediate experience, since each of the four perspectives either stems from it or constitutes its ground. Following the model of the cross [see I.3 and III.3], we then plot on the horizontal axis stretching out from experience in both directions the two perspectives which yield synthetic knowledge: to the right lies a priori knowledge and to left a posteriori knowledge. And on the vertical axis, as it were, cutting into the synthetic axis at the point where it meets experience, we plot the two perspectives which yield analytic knowledge: above experience is the a posteriori belief in a reality which transcends experience, and below it is the abstract a priori knowledge of logic. The manner in which each perspective is connected to experience by a particular sort of reflection can now be represented adequately by making each axis into an arrow. Thus, in the case of transcendental and practical reflection, the arrow points towards experience, since each of these is an attempt to determine the ultimate principles which act as its ground in one way or another; and in the case of empirical and logical reflection the arrow continues in the direction of its counterpart, so that it points out from experience, since in both cases the flow of thought presupposes experience as a basis (either for empirical reasoning or for logical abstraction).
Interpreting Kant's System in accordance with this map enables us more readily to detect the short-sightedness of many interpreters, such as Hintikka, who accuses Kant of arguing that 'we as it were look at the world always from the same perspective' [H17:94], or Allison, who tends to underestimate the importance of the other perspectives, as when he claims 'transcendental reflection ... can be taken as equivalent to the critical method itself.' By the same token, it enables us to grasp the appropriateness of other, potentially confusing ways of interpreting various aspects of Kant's System, such as Paton's description of the first half of Kt1 as a 'metaphysic of experience' [P2]. Although systemt is not technically part of Kant's metaphysics, Figure IV.2 suggests an analogy revealing a sense in which Paton's epithet is appropriate. Just as metaphysical reflection operates at the opposite pole of the same axis as logic, transcendental reflection determines the 'metaphysic' at the opposite pole of the axis of our ordinary empirical reflection on experience.
The correspondence between this fourfold division and that between Kant's four faculties [see II.4 and Figure III.10] is fairly obvious, and its importance will become even more evident in Part Three. When the four stages in each of Kant's systems are viewed in terms of the four perspectives from which experience can be interpreted, it becomes clear that they are not gradually improving 'versions' of the same basic argument, repeated in ever new and updated forms, as are the 'stages' in Wolff's interpretation [W21:111]. Rather, they form a progressive development, a set of cumulative conclusions linking the variety of (often seemingly contradictory) steps leading to knowledge, which combine to constitute a single, unified system.
Although our discussion of Kant's reflective perspectives in IV.3 concentrated mostly on their functions in systemt, we saw in II.4 that each of these can also be used as a standpoint from which to view all the perspectives in a given system. Along these lines, our map of Kant's essential perspectival method can also make it easier to explain how the three Critiques fit into the overall plan of his System of Perspectives. Using the Transcendental as a kind of umbrella Perspective for all the other perspectives, Kant begins his Critical philosophy by using the logical perspective as the basis for the theoretical standpoint in systemt. He then proceeds to the opposite pole in Figure IV.2, and uses the hypothetical perspective as the basis for the practical standpoint in systemp. Finally, he completes the circuit by using the empirical perspective as the basis for the judicial standpoint in systemj. Since 'judgment' for Kant is primarily an empirical activity [see E4:480; G6:457], a 'critique of judgment' is bound to be significantly different from the other two types of critique, whose standpoints it attempts to unite in a common, third standpoint. This difference is appropriately represented in Figure IV.2 by placing judgment's empirical perspective on an altogether different axis than the perspectives of the first two Critiques. Moreover, placing it on the pole opposite to the transcendental perspective suggests the paradoxical nature of the task set for the third Critique: its standpoint is in a sense opposed to, and yet the fulfillment of, the root Perspective of the entire Critical philosophy.
The general picture of Kant's fundamental epistemological perspectives presented in Figure IV.2 is, of course, only one of the preliminary steps towards a coherent interpretation of his System. Fully substantiating my claim that Kant's philosophy is profoundly coherent [see IV.1] will necessitate applying this framework to innumerable problems and ambiguities which arise both in his writings and in those of his interpreters and critics. It will therefore be most suitable to proceed from here to an interpretation of the 'thorny' topic of the transcendental object (which includes notions such as the 'thing in itself', 'appearance', etc. [see Ch. VI]) and from there to the transcendental subject (i.e., the role of intuition, conception, etc. [see Part Three]). We will then be prepared to use this interpretation as a guide to making an accurate assessment of the metaphysical implications of Kant's System [see Part Four]. By constantly keeping in mind the perspectival framework offered in this chapter, our analysis of the elements of this System will perhaps enable it, as Kant hoped, to 'secure for itself the necessary elegance of statement' [Kt1:xliv].
 S17:38-42. Kuehn lists 21 remarks of this type to show how those in 'what may broadly be called the "Strawsonian tradition" of the study of Kant' have (unjustifiably) 'raised to a highly refined art the ascription to Kant of [such ineptitudes]' [K15:512].
 Kant always intends the word 'subject' (as well as 'mind') to be taken 'in a comprehensive sense, as inclusive of all who belong to the human race' [Kt7:401; s.a. C10:186-7]. But its specific function in each context depends on the perspective being adopted.
 Kt1:276-7; s.a. 275,276,864. Of the 48 occurrences of 'immediate' in Kt1 [see Pq10:179], most of the others concern either the 'immediate representation' [Kt1: 41,94(2),A252] of an object in 'immediate intuition' [48,195,273] or sensation -i.e., the 'immediate relation' of one representation to another [33,93,180, 327,363,671,690]-or the experience of 'immediate consciousness' [276,276n, A372; s.a. A354-5,A371], or their combination in an 'immediate perception' [272,273,A368,A371,A377,718]. (Even more examples of each could be cited if the 67 occurrences of 'immediately' were also included in this listing.) Such immediacy is implied by Kant's use of 'experience' in various passages in his early writings, as when he says that the 'reason of knowing' something (the 'ratio quod') is always given in 'experience' [Kt11:392(220)] or that in metaphysics 'concepts derived from experience ... must always be the foundation of all our judgments' [Kt18:367-8(113)]. Kant also uses 'immediate experience' from time to time in his later writings [e.g., Kt7i:240].
 Humphrey refers to Kt1:1-3 and several other texts [A1-2,116-7 and Kt69:274-5] as evidence of Kant's distinction between experience in 'the ordinary sense', as 'an isolated act of perception', and experience in Kant's technical sense, as coextensive with 'theoretical knowledge': the former, he says, concerns 'the questio facti of experience' while the latter concerns 'the questio juris about experience' [H22:26-7]. Walsh distinguishes in a less rigorous way between such 'immediate' experience and the more 'developed' sense of experience, noting that this distinction is reflected in Kant by the difference between 'Empfindung' (feeling or sensation) and 'Wahrnehmung' (perception) [W7:221-2].
 Kt1:74; s.a. Kt5:455. The precise meaning Kant gives these terms in his analysis of the stages through which experience passes will be discussed in VII.1-2. Some of these stages would themselves be called 'experience' in ordinary language (e.g., conscious activity which is not focused on a given object, or the perception of an object which does not enter fully into conscious thought). But Kant would regard such 'experience' as 'merely subjective', and would give it some other name, such as 'imagination' or 'apprehension' [see e.g., Kt1:A115-28].
 Kt1:34-5; cf. 74. That Kant replaces his initial explanation of this distinction with a clearer version in the second edition of Kt1 reveals his increasing awareness of the importance of specifying technical meanings for his primary distinction between knowledge and experience. However, he could have avoided using the word 'empirical' for both his pure-empirical and his empirical-transcendental distinctions [see IV.3] simply by replacing it in the former case with 'non-pure' [cf. C14:246-7] or 'mixed' (as in Kt14 [see e.g., 50(83); cf. Kt1:A11]). Alternatively, it might have been even better for him to have identified pure knowledge with knowledge by reflection-i.e., with knowledge which arises out of thinking about experience, whether or not it requires the intuition of an object in an actual experience. The pure-empirical distinction would then have been one between nonreflective experience (empirical) and knowledge by reflection (pure), much like the one I shall develop in this chapter. Although any type of reflection would be regarded as yielding pure knowledge, different 'levels' of purity would have to be discerned (e.g., transcendental, empirical, etc.). However, Kant does not use his terms in this manner.
 Kt1:165-6; s.a. 147,218; Kt7i:203n. In K2:11.302(Z1:184) Kant describes the way he proceeds in his lectures: 'I begin by defining "experience" in terms of empirical knowledge.' As he puts it in Kt69:274 [s.a. 276]: 'Cognition [Das Erkenntniß] of the objects of the senses as such ... is experience.'
 Kt1:5-6; cf. Kt7:179. As we shall see in IV.3, some reflective knowledge is not strictly pure or empirical, but an 'admixture' of both [Kt1:3].
 A posteriori knowledge can also be called 'necessary', but only when the necessity is derived to some extent from our experience of the laws of nature. For example, it happens to be a necessary truth that human beings cannot survive prolonged exposure to temperatures above, say, 100°C; but this fact has only a posteriori necessity because its truth is discoverable only by reflecting on the structure of the natural world, and not on the laws of thought.
 Nakashima calls attention to this ambiguity [N1:98-9], regarding it as a reason for denying the validity of Kant's overall Transcendental Perspective. Pippin, by contrast, recognizes the importance of actually clarifying Kant's meaning. After stating 'it is not so clear just what this kind of formal knowledge is' [P8:20], he presents an exhaustive account of Kant's theory of form which helps clarify this ambiguity [see e.g., 91,94-5]. He explains at one point that 'A priori does not mean "not derived from experience" but "known without appeal to experience"' [102; s.a. P4:203-4]. Pq8 and Pq9 give a detailed account of the meaning of the phrase 'a priori knowledge'. In Pq8:9-14 I pay special attention to the ambiguity under consideration here.
Werkmeister tries to clarify the nature of a priori knowledge by noting that Kant always 'uses a priori adverbially, not adjectivally' [see e.g., W17:213]. Indeed, he warns against the common practice of using the phrase 'synthetic a priori knowledge' in interpreting Kant [e.g., 66-7,215-6], claiming this 'is perhaps the worst possible misunderstanding of Kant' . Unfortunately, he not only fails to give any clear explanation of the meaning of the term 'a priori', but he also never actually explains what is so bad about this common way of translating Kant's phrase into smooth English. Provided we keep in mind that 'a priori' and 'a posteriori' refer to how we know things, not to any intrinsic qualities, an adjectival translation would not seem to cause any serious problem. For although Kant's normal usage is undoubtedly adverbial, it is rather presumptuous to say it is always such, without citing extensive textual evidence. Moreover, Werkmeister himself uses 'a priori' as an adjective on at least one occasion, where he refers to '... the synthetic and a priori elements ...' ! Yet this need cause no confusion, as long as we remember to associate both 'a priori' and 'a posteriori' with adverbial meaning and both 'analytic' and 'synthetic' with adjectival meaning.
 W21:304. Paton goes so far as to equate Kant's distinction between a priori and a posteriori with his formal-material distinction [P3:61].
 Cf. Figures II.1 and III.9. The fact that Figure III.9 depicts each Critique as filling the position of the transcendental perspective in relation to the other works in its system [cf. Figure IV.2] should clarify the reason for the broad use of words such as 'transcendental' and 'a priori', mentioned here. Unfortunately, it is not always easy to discern which use Kant has in mind in specific contexts. To make matters worse, the term 'pure' is, as mentioned above, often used in roughly the same way as the broad senses of 'a priori' and 'transcendental', though their meanings are technically distinct.
 Although these terms are 'as old as Euclid' [P2:1.130n]-indeed, older [H9:I.136-42]-and although Kant's usage was largely influenced by his predecessors [H4: lv-lxxvii], his peculiar formulation of this distinction is certainly original, as Allison convincingly argues in A11 [s.a. K2:11.38(Z1:141)]. Before Kant, the distinction referred not so much to the epistemological status of judgments, as to reciprocal methods of argumentation [see Ap. IV and Pq18:I.3]. In H18 Hintikka examines in detail the roots of 'the method of analysis and synthesis' in ancient Greek geometry.
It should also be noted that, according to Kant, 'knowledge' always reveals itself in the form of 'judgments', not just 'propositions'. When he speaks, for example, of 'analytic judgments', he must be taken to mean roughly the same thing as when he speaks of 'analytic knowledge' [cf. Kt1:10 and 191]. Thus, although Wolff is, at least in general, grammatically correct to point out that '"analytic" and "synthetic" are adjectives which modify the noun "judgment," while "a priori" and "a posteriori" are adverbs which modify the verb "know" and its cognates' [W21:113n; see note IV.10], the two sets of terms nevertheless have the same field of applicaion. The empirical connotations of the word 'judgment' [see K9: 18 and II.4] should not, as we shall see in Appendix IV, mislead us into limiting the analytic-synthetic distinction to an empirical interpretation.
 Kt69:32. Kant appears to contradict this claim when he says 'the law of [non]contradiction ... is absolutely incapable of grounding any but identical judgments' [K2:11.36(Z1:139)]. He clarifies the relation between analytic and identical judgments in Kt69:322: 'analytic judgments are not identical because they require division and in such division serve to clarify their concept, whereas, by contrast, in identical judgments ... absolutely nothing is explained.' Humphrey explains that this means analytic judgments, though reducible to identity, may be able to clarify concepts [H22:22; s.a. Kt10:111(117)].
 Kt22:241; s.a. Kt1:749; Kt10:111(118); A7:60,164; A11:36-7. Kant refers to this as 'the principle of synthetic judgments' [e.g., Kt22:241 and K2:11.38(Z1: 141)], thus highlighting its direct correlation with the principle of noncontradiction for analytic judgments. On the nature of an 'intuition', see VII.2.A.
 B5:5,10; cf. Kt1:A8. Kant gives his own formulae in Kt10:111(117-8): 'To every x to which appertains the concept of [the subject] (a+b) appertains also [the predicate] (b)-is an example of an analytic proposition. To every x to which appertains the concept of [the subject] (a+b) appertains also [the predicate] (c)-is an example of a synthetic proposition.' Allison explains that the latter connection between 'the predicate (c)' and 'the subject concept (a+b) ... is grounded in ... the reference of both to the identical object (x)', whereas for an analytic judgment, the predicate simply reiterates something already 'contained' in the subject concept without requiring any 'extra-conceptual claim' about 'the "reality" of the predicate' [A11:36-7].
 This holds, of course, only as long as the meanings of the words involved are already understood [see H1:108-9 and S20:266-7,231]. For someone who does not know what 'yellow' means, the proposition 'Yellow is a color' would be informative; but, as I have argued in Pq9:273-9, it would not then be analytic in the usual a priori sense for that person. This will become more clear when we discuss the possibility of analytic a posteriori knowledge [see IV.3 and Ap. IV]. But in either case, Garver is right when he says 'the clarification achieved through analytic propositions consists in presenting immediate inference possibilities pertaining to some word which expresses the concept that is being clarified' [G3:266; s.a. K9:19; H4:xxxix-xl; P4:136-7].
The inevitable existence of human ignorance renders inadequate those contemporary accounts of analytic knowledge in which it is equated with 'deducible from definition' [B5:19; see e.g., W1:31]. For as Beck aptly insists, Kant regards definition as 'a sufficient, but ... not a necessary, condition for analytic judgments' because it 'requires a completeness and precision that is often an unattainable ideal; yet its absence does not jeopardize the analytic judgments already made' [B6:34]. Therefore, 'a judgment logically implied by a definition is analytic, [yet] analytical judgments are not necessarily or even usually known or justified from definitions' [36; cf. Kt10:140-5(141-6) and H4:xxii-liv].
 Although it is in virtue of the laws of formal logic that the informative content of analytic judgments is reducible to nothing [Kt1:190; cf. B5:19], it should be stressed that the analytic-synthetic distinction in general 'is not one of formal logic, for formal logic abstracts from the meaning of all terms' [10-1]. This is the point Zweig is making when he urges it is wrong to think Kant says 'analytic judgments are deducible from the principle [of noncontradiction] alone': for Kant regards this principle 'as a rule to be used in testing a judgment and not as a premise from which other propositions are to be deduced' [Z2:167]. However, too much emphasis on the 'transcendental, non-logical nature of the analytic-synthetic distinction' [as in A7:59; s.a. 46-75] can be misleading, since, as I shall argue in IV.3, logical reflection is also a necessary requirement for the determination of analyticity in its Kantian sense. The distinction itself arises out of the relation between the transcendental and the logical perspectives, so it cannot accurately be described in terms of one or the other on its own.
 The debate was first formulated in these terms by Quine, who argued that in the analytic-synthetic distinction the entire 'difference is only one of degree' [Q1:43]. Subsequent arguments for and against this position have been too numerous to review here, though a sampling of these developments is presented in Appendix IV.
 Hartman and Schwarz describe the 'analytic a posteriori' in terms of 'empirically analytic' qualities which 'are part of the experience of the thing', yet are 'merely perceived but not yet conceived' [H4:l-li]. And Cameron suggests the proposition 'I have experience' as a possible candidate for this status [C2:352-3; s.a. M7:34-42]. I argue towards the end of Appendix IV, and more thoroughly in Pq9, that what recent philosophers such as Kitcher and Kripke have called the 'contingent a priori' is more accurately regarded as another example of the analytic a posteriori.
 Scruton's paraphrase of this question ('How can I come to know the world through pure reflection, without recourse to experience?' [S8:19]) is misleading for two reasons: first, because (as we shall see in IV.3) all forms of reflection are related in one way or another to immediate experience; and second, because synthetic a priori knowledge is concerned not with the world as such, but with the forms imposed on it by the knowing subject. Prichard's paraphrase is more appropriate: 'How is it possible that the mind is able, in virtue of its own powers, to make universal and necessary judgments which anticipate its experience of objects?' [P14:33; s.a. 19; cf. A7:2; E3:83; G10:152; P4:135].
 V2:445. Calling reflective judgment subjective is not intended to degrade it, but only to distinguish its 'inner' character from the 'outer' character of ordinary experience. However, van de Pitte misleadingly argues that determinant judgment depends on reflective judgment . A theory of determinant judgment is indeed a product of reflective judgment; but this does not require a person to be aware of even the possibility of reflection before employing determinant judgment [see below]. As Wallace puts it: 'The reflective judgment ... looks at the relation between the mental representation of the object [via determinant judgment] and the general constitution of the human mind' [W5:192].
 That is, empirical reflection focuses on what Kant calls 'phenomena' [Kt1:A248-9,306; see VI.3]. Kant clarifies the meaning of 'empirical knowledge' in K2: 11.302(Z1:184) by saying a representation is empirical if 'the object is given in a sensuous representation (which ... includes sensation and ... consciousness ...)' and it is knowledge if the perceived object 'is thought' by means of a 'conception'. Calling this synthesis of intuition and conception 'experience' highlights the fact that, as we shall see in VII.2.A, intuition on its own cannot generate a properly empirical perspective.
 Webb neglects Kant's sincere respect for the viewpoint of the common man [see XII.1] when he claims in W13:211 that an 'experience in which the whole of our personality is involved' is for Kant 'somehow inferior in validity to the results of abstraction.' On the contrary, such an experience represents a different, but equally valid perspective, even though some perspectives are more appropriate for the philosopher to adopt while asking certain types of questions. Thus for example, Kant's contrast between 'the empirical' (or 'historical') and 'the rational' (i.e., 'the whole higher faculty of knowledge') in Kt1:863-4 is not intended to diminish the validity of the former, but simply to define a difference between two Perspectives: 'Historical knowledge is cognitio ex datis; rational knowledge is cognitio ex principiis.' The former is knowledge either 'through immediate experience' or 'through instruction', while the latter is knowledge either 'from concepts ['philosophical'] or from construction of concepts [mathematical]' [864-5].
 Kt1:316-49. Stressing the significance of transcendental reflection, Allison goes so far as to say it 'can be taken as equivalent to the critical method itself. Consequently, ... the errors of all non-critical philosophies are traceable to a failure to engage in transcendental reflection' [A10:45]. He neglects to point out, however, that this common failure is itself a direct consequence of failing to draw the distinction between the transcendental and empirical perspectives in general-a distinction which he does emphasize in A6. Kant himself stresses the importance of this distinction in the very title of the Appendix, which states that 'The Amphiboly of the Concepts of Reflection' arises 'from the Confusion of the Empirical with the Transcendental Perspective' [Kt1:316; s.a. 345-6]. The disastrous consequences of failing to take into account the perspectival nature of this distinction are evident in many interpretations [s.e. P14].
 Kt1:A96. Propositions are transcendental if, as Ewing puts it, they can be 'proved by showing that if they were not true of objects, these objects could not be experienced by us' [E5:26]. The nature and function of Kant's special type of 'transcendental arguments' will be discussed in V.3 and in Appendix V.
 See IV.2. Such reflective experience contains certain aspects which can be traced back to a nonempirical source. In itself-i.e., before the philosopher actually comes to know it in transcendental reflection-the knowledge revealed in this 'tracing back' is not really 'knowledge' at all, but the necessary condition for the possibility of both reflective and determinant judgment, which every knowing subject naturally follows unconsciously. Kant's reference to such conditions as 'knowledge' has given rise to various misunderstandings of what he means-misunderstandings which usually lead to a premature rejection of his views [see e.g., my criticism of Walker in Appendix V.3 and of Kitcher in Pq8:9-14].
 See Kt1:319. Unfortunately, aside from explaining what he means by 'transcendental', and arguing for the validity of various synthetic a priori knowledge-claims, Kant never gives a detailed explanation of how it is possible for human beings to attain such knowledge. This has made it easier for some philosophers to reject its legitimacy; but Walsh defends its possibility admirably, in my opinion, in W9:249-59.
 Kt1:318-9. The neglect of the crucial difference between these two perspectives is one of the central contentions of Kant's polemic with Eberhard [see e.g., Kt22:193-4].
 See e.g., Kt22:221,244-5; A7:164; S7:171. Along these lines, Neeman suggests in N2:8-9 'that the distinction between analytic and synthetic propositions rests for Kant on two different kinds of real acts of cognition. For the logical nature of the object is supposed to be discovered analytically and its real nature synthetically.'
 See Ap. IV. Because logical reflection is ultimately rooted in experience [Kt1: 195], it is, of course, dependent on linguistic conventions [see note IV.17]. But any analysis of these conventions as such is empirical, and has no part in determining a proposition's logical status as analytic or synthetic. Accordingly, Weitz's assertion that analyticity applies only to 'a statement which [merely] expresses part of the everyday usage of the term[s]' [W15:492] should be regarded as a variant of Quine's empirical perspective on the distinction [see note IV.19].
 Kt1:xxi; s.a. x,xxvin and Kt8:181(169). In Kt1:661 Kant defines 'practical knowledge as the representation of what ought to be.'
 Kant rarely uses the term 'reflection' in close connection with the terms 'hypothetical' or 'practical' [but see Kt39:476(182)]. Nevertheless, these are legitimate names for this type of philosophical reflection, because he does connect the closely related Metaphysical Perspective with reflection [e.g., Kt1:366-7], and he frequently uses phrases which are equivalent to 'hypothetical perspective' [e.g., 675] and 'practical standpoint' [e.g., xxi,384]. (My reasons for treating 'hypothetical' and 'practical' as more or less synonymous terms are given in II.4 and Appendix II; see also V.4.)
 A8:238; cf. Kt1:599-601. Unfortunately, as we shall see, Kant does not clearly recognize the implications of this fact.
 Kt1:799. Kant's reason for occasionally calling the ideas 'fictions' is not because he thinks they really are entirely fictitious [see notes V.9,10], but simply to guard against the dogmatic mistake of assuming 'their independent authority or absolute validity', which would run the risk of 'drown[ing] reason in fictions and delusions' . Kant's point is that the reality of the ideas is demonstrated not in systemt, but in systemp [see Kt4:5].
 See note IV.37. As we shall see in VIII.2-3, one of the main tasks of the practical standpoint is to discover the implications of the 'categorical imperative'. Thus, what is 'hypothetical' for theoretical reason becomes 'categorical' for practical reason. But this should not obscure the fact that the fundamental perspective which guides systemp is the same one which in systemt is called hypothetical. This simply means the difference between hypothetical and categorical is, for Kant, a difference of standpoint.
 Kt1:830; Kt5:388. Although it could be said that the knowledge yielded by practical reflection in systemp makes moral experience possible, it would be more accurate to say it makes moral experience coherent, or rational, by providing its justification, and that it consists in principles which ought to be applied universally to experience.
 The role of hypotheses in the formation of faith will be discussed further in V.1 [s.e. note V.6], and the relation between faith and these ideas will be discussed in V.4.
 Kt4:134. Statements such as this by Kant reveal the inadequacy of labeling the product of practical reflection 'synthetic a priori'. The difference between 'knowledge' and 'belief' will be discussed more fully in V.1 and Appendix V.3.
 And the same goal is aimed at in systemp, wherein the hypothetical perspective gives rise to the practical standpoint, which provides a more positive alternative to the speculative perspective by reflecting on what 'ought to be true' in light of the universal experience of duty [see VIII.3.A; s.a. Kt1:700-1].
 That it is appropriate to locate logic below experience with the arrow of reflection pointing down towards it is intimated by Kant when he says logic always comes 'last of all' in the actual formulation (as opposed to the logical structure [cf. III.2-4]) of any science [Kt1:76].
 That it is appropriate for synthetic a priori knowledge to 'point to' experience in this way is suggested by a metaphor of Kant's, according to which 'pure a priori concepts ... must be in a position to show a certificate of birth other than that of descent from experiences' [Kt1:119e.a]. In Kt18:358(98) he uses another metaphor which also alludes to this pattern: 'knowledge has two ends of which you can take hold, the one a priori, the other a posteriori.'
 A10:45. Of course, in its broadest sense 'transcendental' does refer to the Perspective of Kant's entire Critical philosophy [see II.4]; yet, as we saw in II.2, the Critical method should be associated not so much with any particular perspective as with Kant's natural tendency to think perspectivally, especially since he was thinking Critically long before he hit upon the insight which led to his new Transcendental Perspective [s.a. Pq12]. In fairness to Allison, however, it should be noted that Kant himself often seems to underestimate the philosophical significance of empirical reflection. (By contrast, most philosophers would now agree that the only way they could fulfill Kant's hope [Kt1:viif] of making philosophy (excluding logic) into a 'science' (at least, the way this term is used today) would be to adopt some type of Empirical Perspective-such as, perhaps, some version of verificationism or linguistic analysis.) Nevertheless, even though Kant stresses the importance of one perspective more than the others, they are all equally essential to a coherent understanding of his overall method.
Incidentally, Kant's emphasis on the transcendental led him, much to the dismay of Husserl and numerous others, to be satisfied with a fairly uncritical view of logic 'as self-sufficiently grounded in its apriority' [S12:49]. Kant would not deny the legitimacy of investigating the foundations of and the justification for logical reflection as such (rather than as speculation), but this task is secondary to the properly Transcendental tasks Kant undertakes in his three Critiques.
 See Kt1:682-3. Ordinary metaphysics regards logic as applicable directly to experience, and therefore uses it to make 'empirical-like' inferences about transcendent reality. This can be plotted onto our map of Kant's methodology by reversing both vertical arrows in Figure IV.2 (as in Figure III.8). Kant's discovery was to recognize that such applications must be preceded by Transcendental Criticism if they are to be properly made.
 As mentioned in III.3, this way of organizing Kant's System follows the same pattern as his division of categories [cf. Figures III.4 and IV.2].
 Kt1:194-5. Accordingly, as long as it involves judgments which synthesize the theoretical and practical standpoints, any class of experiences-not just the aesthetic and teleological-could be chosen for this task.
Send comments to the author: StevePq@hkbu.edu.hk
Back to the Table of Contents for this book.
Back to the listing of Steve Palmquist's published books.
Back to the main map of Steve Palmquist's web site.