Kant's Categories Reconsidered

Robert Greenberg

Brandeis University

ABSTRACT: Adopting a Quinean criterion of ontological commitment, I consider the question of the ontological commitment of Kant's theory of our a priori knowledge of objects. Its direct concern is the customary view that the ontology of Kant's theory of knowledge in general, whether a priori or empirical, must be thought in terms of the a priori conditions or representations of space, time, and the categories. Accordingly, this view is accompanied by the customary interpretation of ontology as consisting of Kantian "appearances" or "empirical objects." I argue against this view and interpretation. My argument turns on the opposition between the necessity and universality of the a priori and the particularity and contingency of the existent. Its main point is that the a priori can remain necessary and universal only if the existence of objects is kept distinct from it.

I. Introduction

To the extent that category theory, i.e. that there are certain predicates of things that are fundamental to our thought about objects in general, has been based on our thought of objects of possible experience, it has been highly suspect. This is the negative thesis of this paper. Over the years, philosophical inventiveness has produced various schemes of predicates which challenge the claims of necessity that have been made on behalf of the scheme we employ for such objects-a scheme of substances that are involved in causal action and interaction. If no particular scheme is necessary, perhaps it is not necessary that we employ any scheme at all.

Kant's theory of categories is no different from any other category theory in this regard. Its dependence on what Kant calls the logical functions of judgment doesn't change matters. If it is primarily experience that the categories are supposed to help make possible, these functions are indistinguishable from logical forms of of judgment, and such forms, belonging as they do to formal, or what Kant calls general, logic, a logic that abstracts from all content of knowledge in general, cannot instruct us with respect to fundamental predicates of things.

This negative thesis suggests a positive one, namely, that categories should be argued for as necessary conditions of the possibility of a knowledge that is distinct from experience, viz., a priori knowledge. It is only in the context of such an argument that the categories can be shown to be required for possible experience as well.

Before we turn our attention to our separate arguments in support of the two theses which we are attributing to Kant, we should first consider an objection of interpretation that applies to both attributions. It might be noted that for Kant the objective validity of the categories is not completely established until they are shown to be necessary conditions of the possibility of experience (e.g. B 161). (1) It might then be asked why the demonstration of this necessity cannot be separated from that necessity pertaining to the possibility of a priori knowledge, as Strawson (2) and Guyer (3) recommend.

The answer is that in demonstrating that the categories are necessary for possible experience Kant takes the objects of possible experience to be appearances (e.g. B 159 ff.). However, as Guyer himself would be the first to agree, Kant's appearances are intelligible only in a context in which it is a priori knowledge, not experience, whose possibility is under discussion. Therefore, that is the context in which the categories are also shown to be necessary for possible experience. That is, apart from that context, Kant has no argument for the view that the categories are necessary for possible experience. All this, of course, is predicated on the proposition that Kant's appearances are to be understood in terms of his theory of the possibility of a priori, not empirical, knowledge-admittedly a proposition only some of us subscribe to, yet, unfortunately, one that can only be argued for elsewhere.

We are now in a position to argue for the two theses separately.

II. The Negative Thesis

Kant himself disposed of his first, or A-, edition transcendental deduction of the categories on the grounds of its "obscurity" (B xxxviii). But it may very well be that the argument contains a more serious problem than that.

The question confronting the transcendental deduction, let us recall, is whether appearances, given by means of sensibility, must be subject to the conditions of the unity of the distinct faculty of the understanding (A 89-90/B 122-3). In the A-Deduction Kant thought he could answer the question affirmatively, if, having already (in the metaphysical deduction) identified the categories as the conditions of the unity of the understanding , he merely took the extra steps of (1) identifying that unity with the transcendental unity of apperception and (2) laying down that unity as the principle that all objects must be subject to if they are to be taken up into the necessary unity of knowledge (A 103 ff.). So the transcendental unity of apperception is the principle of the possibility of all objects with respect to our knowledge of them, i.e. the epistemic possibility of objects, insofar as that knowledge contains the necessary unity of thought.

One would expect, therefore, that in advancing beyond the metaphysical deduction to prove that all appearances must be subject to the categories if they are to be known, the pivotal item in that advance-the transcendental unity of apperception-would thereby be explicitly shown to be connected to the categories.

But that isn't done in the A-Deduction. Rather, Kant merely asserts that he "maintain[s] that the categories . . . are nothing but the conditions of thought in a possible experience" (A 111). Though he has previously provided an argument that the transcendental unity of apperception depends on "rules" governing the "reproduction of the manifold" of intuition (A 105), the burden of specifying which rules at bottom they must be falls entirely on the metaphysical deduction (Cf. A 119, A 125). In other words, although the argument has advanced beyond the metaphysical deduction through the introduction of the transcendental unity of apperception, the categories haven't keep pace.

Now, it is no coincidence that so many of those who criticize Kant for failing to bring the categories up to the distinctive demands of the transcendental deduction brought about by the introduction of the transcendental unity of apperception into the argument are commentators who are actually sympathetic to what I consider to be the unfortunate method which the A-Deduction actually seems to follow. This is the method of trying to prove the necessity of the transcendental unity of apperception on the grounds that it is a presupposition of the possibility of experience. That is, they would be among the first to notice that since Kant is arguing for its necessity on these grounds, he needs further argument that would serve to connect it to the categories on the additional grounds that the categories themselves are necessary for the possibility of experience. And these further grounds are provided, they claim, only later on, in the Analytic of Principles. The commentators counted among these are legion. (4)

To see what it would take to upgrade certain propositions in the metaphysical deduction so that they can adequately serve as premises in the transcendental deduction, we should first notice that it is following the metaphysical deduction that Kant raises the question facing the transcendental deduction in terms of "the functions of the understanding" (A 89/B 122) or "the functions of thought" (A 91/B 123), as well as in terms of the categories (A 85/B 117, A 88/B 120). That is, he refers to the conditions of the unity of the understanding as such functions as well as referring to them as the concepts that "rest on" the functions, i.e. the categories (A 68/B 93). It is therefore evident that the functions referred to are the logical functions of judgment, since the metaphysical deduction has just tried to establish that these are the functions that stand in isomorphic relation to the categories and that "specify the understanding completely, and yield an exhaustive inventory of its powers" (A 79/B 105).

So let this isomorphism between the categories and the logical functions of judgment be our clue toward discovering what must be done to make the premises given by the metaphysical deduction suitable for the transcendental deduction and its distinctive reliance on the transcendental unity of apperception. Since the grounds of this isomorphism consist in two identity-propositions that are crucial to both the metaphysical and the transcendental deduction, let us identify them here. The first is:

The same function which gives unity to the various representations in a judgment also gives unity to the mere synthesis of representations in an intuition, and this unity, in its most general expression, we entitle the pure concept of the understanding. [A 79/B 104-5]

Kant is thus asserting an identity between the function (i.e. "the unity of the act [i.e. action [Handlung]] of bringing various representations under one common representation" [A 68/B 93]) that is the source of the unity in a judgment and the function that is the source of the unity of a synthesis in an intuition. (5) We thus look forward to further argument for an explanation of this identity between the source of the unity in a judgment and that of the unity of a synthesis in an intuition; that is, an explanation of the relation between a unitary judgment and a unitary synthesis in an intuition.

The other identity Kant asserts is between the two actions (Handlungen) that contain the just mentioned identical function:

The same understanding, through the same operations [i.e. actions [Handlungen]] by which in concepts, by means of analytic unity, it produced the logical form of a judgment, also introduces a transcendental content into its representations, by means of the synthetic unity of the manifold in intution in general. [A 79/B 105]

Now, it seems universally agreed among commentators that the expression, "the logical form of a judgment," as Kant employs it here, is a generic term for the logical forms of judgment that are given, it is further assumed, by the Table of Judgments, which has been laid down earlier in the Transcendental Logic (A 70/B 95). In line with this reading of the passage is the further reading that the transcendental content Kant is referring to here consists in a synthetic unity of the manifold of intuition in general that is independent of these logical functions (i.e. forms) of judgment.

Both readings, however, are dubious. Presumably, the expression in the passage, "its representations," refers to the categories. One reading would thus have the content of the categories be independent of the logical functions of judgment. But this can't be, since it is through these "functions of unity in a judgment" (A 69/B 94) that "concepts, as predicates of possible judgments, relate to some representation of a not yet determined object" (ibid.). And surely we can understand this relation of concept to object as the content of the concept (See A 55/B 79). So, for this reason, as well as others, (6) we might rather view the transcendental content in question as belonging to the logical functions of judgment. (7)

Where does this leave Kant's reference to the logical form of a judgment? A clue can be taken from the B-Deduction. There, this form is identified as consisting in "original apperception and its necessary unity " (B 142). As a principle that gives rise to "the objective determination of all representations" (ibid), it is none other than "the transcendental unity of apperception" (ibid.). But it is the necessity of this unity that is the very issue before us! It is with respect to this unity that we are looking to an upgrading of the identity-propositions from the metaphysical deduction for the conditions under which objects or representations may be brought to it.

Well, the passage from the metaphysical deduction we are presently reading tells us that it is the same action that produces the transcendental unity of apperception in concepts that also introduces the content of the logical functions of judgment into the categories. So, since the issue before us is how, independently of the Analytic of Principles, Kant can support his thesis that the categories are the conditions under which objects or representations may be brought to the transcendental unity of apperception, we have only to look for an argument that spells out the relation between the logical form of a judgment that is produced in concepts and the transcendental content that is introduced into the categories. But, under the present proposal, this is a relation between the transcendental unity of apperception and the content of the logical functions of judgment. Furthermore, Kant asserts that the relation obtains through an action that is identical to both of them. It is thus an argument that establishes a relation between this form and this content through an identical action that is involved in each. It is precisely this relation through this identical action (which Kant entitles combination [B 130 ff.]) that finally gets spelled out in the conclusion of the "first step" of the B-Deduction, i.e. in Sections 19 and 20 of that deduction. (8) And, necessarily, this is done as the B-Deduction also explains the identity asserted in the first identity-proposition considered above, namely, the identity between the unity in a judgment and that of a synthesis in an intuition. Of course, this identical unity is none other than the transcendental unity of apperception itself. In a word, these two identity-propositions from the metaphysical deduction set the initial tasks for the transcendental deduction.

It is precisely these explanations that are missing from the A-Deduction. By sliding over the need for such explanations, the A-Deduction fails to tell us how the transcendental unity of apperception depends on the categories for objects or representations to be brought to it.

III. The Positive Thesis

A priori knowledge alone is the sort of knowledge that allows the readings of these two identity-propositions from the metaphysical deduction that are required for a successful transcendental deduction. The logical functions of judgment can have the content attributed to them on the proposed reading of the second identity-proposition if and only if the discussion of these functions which surrounds the Table of Judgments belongs to transcendental, not formal, or general, logic (A 55/B 79). Moreover, logic is thus transcendental if and only if any knowledge of which it is the logic is a priori knowledge (A 55/ B 80). Therefore, the logical functions of judgment have the content required for the proposed reading of the second identity-proposition just in case any knowledge that contains them is a priori knowledge. Consequently, they can contribute to the content of empirical knowledge only in the context of their contribution to the content of a priori knowledge.

The proposed reading of the first identity-proposition is also affected, accordingly. We have already proposed that that identical function that gives unity to a judgment and to a synthesis in an intuition, i.e. the transcendental unity of apperception, is the very unity of the identical action, i.e. combination, that is involved in the second identity-proposition. This combination is none other than that involved in a judgment. " . . . we can reduce all acts of the understanding to judgments" (A 69/B 94). And " . . . the understanding . . . is nothing but the faculty of combining a priori, and of bringing the manifold of given represetations under the unity of apperception" (B 135). So, the unity of the identical action in the second identity-proposition is the very unity that is referred to in the first identity-proposition. Of course this must be the case, since a judgment and a synthesis in an intuition are both combinations (B 135, B 142, B 143, B 150, B 162b) and the unity belonging to all combinations is the transcendental unity of apperception (B 131, B 132, B 134-5).

It might be objected that the structure of the proposed interpretation of the B-Deduction doesn't follow the actual structure of the Deduction itself. The interpretation seems to tie the triadic interrelation among the categories, the logical functions of judgment, and the transcendental unity of apperception too closely to a priori knowledge, when, in fact, that interrelation is laid down in the first step of the Deduction quite independently, not just of experience, but of a priori knowledge as well. Worse, a priori knowledge of mathematics aside (B 154-5), the possibility of a priori knowledge of the laws of nature in general is established in step-two (B 164-5) only after, and indeed, on the basis of, the proof of the possibility of experience (B 161). So, either the interrelation in question is established independently of knowledge altogether (step-one), or its involvement in a priori knowledge (in step-two) does indeed depend on first proving its necessity with respect to experience, which is precisely the proposition that is being contested in this paper.

In reply, though the logical functions of judgment are not yet contained in any knowledge in step-one, were that step not to belong to a demonstration of the possibility of a priori knowledge, it would not belong to a logic that is transcendental. This would locate it in formal, or general, logic, if in any logic at all. But that would confound the employment of the categories and the transcendental unity of apperception, not to mention the logical functions of judgment, in step-one. Surely, they contribute to the relation knowledge has to objects; yet that relation is of no concern to formal, or general, logic whatsoever (A 55/B 79).

The objection that the reference to the possibility of experience precedes the reference to the possibility of a priori knowledge of the laws of nature in general takes us back to the beginning of this paper. As noted there, the categories are shown to be necessary for the possibility of experience only in a context in which the objects of possible experience are taken as appearances. But this implies that it is actually the possibility of a priori knowledge that is being proved in that context. Consequently, the proof of the necessity of the categories for possible experience rather belongs to that context and does not stand alone apart from it.

In conclusion, to the extent that the above considerations extend beyond the confines of Kant scholarship, the prospects for category theory generally might be improved, provided the claims made on its behalf are restricted to the possibility that belongs to a priori knowledge alone.

Notes

(1) References to Kant's Critique of Pure Reason are made in the standard way, distinguishing between the first (A/) and second (B/) edition pagination. All translations are from the Norman Kemp Smith translation (London: Macmillan, 1929), unless otherwise noted.

(2) P. F. Strawson, The Bounds of Sense (London: Methuen & Co., 1966), p.86 ff. et passim

(3) Paul Guyer, Kant and the Claims of Knowledge (Cambridge: Cambridge University Press, 1987), p. 73 ff. et passim.

(4) A sample of such commentators might include Jonathan Bennett, Kant's Analytic (Cambridge: Cambridge University Press, 1966), pp. 94-5; Patricia Kitcher, Kant's Transcendental Psychology (New York: Oxford University Press, 1990), p. 89; and Paul Guyer, Kant and the Claims of Knowledge (Cambridge: Cambridge University Press, 1987), pp. 98-9, and "The Transcendental Deduction of the Categories," in Paul Guyer, ed., The Cambridge Companion to Kant (Cambridge: Cambridge University Press, 1992), p 127.

(5) Moving back just a bit in the metaphysical deduction, we find that the pure concept of the understanding is given by "Pure synthesis" (A 78/B 104). So, the same function that gives unity in a judgment and in an intuition can be identified with the previously mentioned pure synthesis. Therefore, it is pure, not empirical, synthesis that gives unity to the representations both in a judgment and in an intuition. This should eliminate any role for the synthesis of the "reproduction of the manifold" in an explanation of the common source of unity in a judgment and an intuition, since that synthesis "is merely empirical" (A 121; see also B 152). But that is part of the explanation of the unity of the manifold of intuition which we find in the A-Deduction (A 105, A 121). So the A-Deduction is already at odds with the metaphysical deduction.

These concluding remarks may prove perplexing to readers who are recalling that being empirical did not prevent the synthesis of apprehension from being carried over from the A- to the B-Deduction (B 160, B 162n.). So why should it be responsible for eliminating the reproduction of the manifold of intuition from the B-Deduction? The answer is that it is not merely its being empirical, but rather its being merely empirical that eliminates the synthesis of reproduction from the argument. The difference between the two is that the concepts, even the necessary ones, i.e. the laws, governing the merely empirical synthesis of reproduction are themselves empirical (A 100, A 121, B 152), even though they must rest on a priori grounds (A 101-2, A 122), whereas the forms to which the empirical synthesis of apprehension must conform are a priori (vide A 99, B 160). The problem with the merely empirical synthesis of reproduction is that what immediately conditions it is itself empirical, and thus psychological (B 139-40). The transcendental deduction must be rid of all psychological conditions on synthesis, including all associative ones, if it is to establish the possibility of a priori knowledge. The B-Deduction does just that, even though it holds over from the A-Deduction the empirical synthesis of apprehension. This empirical synthesis is not a psychological phenomenon.

This position is at odds with the entire approach to the Critique, including, therefore, the transcendental deduction, of Patricia Kitcher in her Kant's Transcendental Psychology (New York: Oxford University Press, 1990).

(6) See Richard E. Aquila, Representational Mind (Bloomington: Indiana University Press, 1983), p. 138, for an inclination suggesting the proposal under discussion. For an actual development of the proposal see Robert Greenberg, "The Content of Kant's Logical Functions of Judgment," History of Philosophy Quarterly 11 (1994): 375-92.

(7) This interpretation of "transcendental content" seems to dispute that given by Darrell Johnson, viz., that it refers to the concept of an object in general. See his, "Kant's Metaphysical Deduction," Proceedings of the Eighth International Kant Congress (Milwaukee: Marquette University Press, 1995) Volume II, Part I, p 273.

(8) The by now widely accepted division of the B-Deduction into two steps was first introduced into the current commentary on the deduction by Dieter Henrich in his, "The Proof Structure of Kant's Transcendental Deduction," Review of Metaphysics 22 (1969): 640-59, reprinted in Ralph C. S. Walker, ed. Kant on Pure Reason (Oxford: Oxford University Press, 1982).