There is little in Manfred Kuehn's account that I disagree with, and despite some misunderstandings I believe that there is a strong convergence in our positions. Manfred Kuehn's title indicates two different, but related, aspects of Kant's account of appearances and things in themselves. There are questions about ‘how’ we might come to think of things in themselves, and questions about ‘why’ we might endorse, in some way, an appeal to such things. In the former Manfred Kuehn thinks one answer, from the ‘absurdity’ argument (Bxxvi–xxvii), is demonstrably weak; in the latter he thinks that I do not do justice to Kant's commitment to a world of things in themselves, especially in the moral philosophy. These are the two topics I consider.
3.1 Appearances and Things in Themselves: How we Come to Think of the Latter
There are some misunderstandings here. Manfred Kuehn asks whether my criticism of the conventional ‘two worlds/two aspects’ labels is directed against Kant or against his commentators. I had hoped to make plain (e.g. 550) that in my view the misuse of that contrast marks errors among commentators rather than in Kant. That misuse has at least two aspects: that the contrast is used, first, without proper clarification of its sense, and, second, with the (mistaken) tacit assumption that we have to choose between the two options. In the first it may not be made clear what the commitments of the alternative views are. If we ascribe a ‘two worlds’ view to Kant does this commit him to the actual existence of things in themselves or only to their conceivability? Should it not be made clear that even if Kant accepts only the conceivability of ‘two worlds’ it is nevertheless two worlds that are conceivable? In the second it may not be recognized that, as Manfred Kuehn and I both point out, Kant demonstrably uses both conceptions. These are standard examples of a careless appeal to conventional labels but they are enough to show how misleading and erroneous such appeals can be. My target is that reference to what have sometimes become, in the literature, thoughtless labels; and it is those thoughtless labels that should be dropped.
My account of the ‘absurdity’ argument provided the opportunity to underline those errors. In Kant it provides an account not only of how we do come to think of things in themselves but also of how we must come to think of them. Formally the argument does not contain any further indication of more positive uses of that necessary conception although Manfred Kuehn is quite right to insist that such positive uses are indicated in what follows (Bxxvii–Bxxx). Manfred Kuehn regards the absurdity argument as ‘very weak’ primarily because it rests on a ‘grammatical’ point about the terms ‘appearance/appears’, and then the question arises: why should we use the term ‘appearance’ at all? He is right that the central point is ‘grammatical’, but I think his objections can be answered.
The first thing is to insist that the argument does not entail commitment to the actual existence of things in themselves (and I think Manfred Kuehn agrees). The second point is to relate the (grammatical) argument to Kant's underlying distinction between sensibility and understanding. Although the point is put grammatically in the B Preface it can be made just as well, as it had been made in A, through the distinction between sensibility and understanding. For Kant ‘appearances’ are supposed, strictly, to belong to sensibility without any reference to the descriptive (conceptual) resources of the understanding. This requirement, artificial and theoretical as it may seem, is captured in his account of appearances as ‘indeterminate’ objects of intuition (B34), which points inevitably to the need to explain that further differentiation into objects which appearances are appearances ‘of’. The term Kant uses to mark the need, in this passage and elsewhere, is ‘thing in general’ (Bxxvii), while ‘transcendental object = X’ has a similar role in the Deduction (A104–6, A109). Kant's subsequent point is that these pressures, whether grammatical or not, can lead philosophers in two radically different directions. In one it leads to a conception of objects wholly distinct from, independent of, our senses and accessed via concepts alone through understanding or reason; in the other it leads to a conception of objects accessed by sensibility and differentiated, on that assumption, through the conceptual resources of understanding or language. The grammatical pressures in the argument acknowledge both the necessity of an appeal to that differentiation into objects, but also the potential error of taking such objects as wholly independent of the senses.
Such an account trades on Kant's later treatment of this underlying philosophical error in the Dialectic and in the Logic (B86–8, B186–7, B342–6). It trades also on his discussion of the Aesthetic in the General Observation where Kant distinguishes the colloquial and the philosophical contexts of talk about things and the way they appear (B62–3, B69–70). In those passages the lesson is that whereas colloquially we are entitled to admit the actual existence of things that appear to us (empirical objects of experience), in the philosophical context we cannot do so if those things are taken to be independent of sensibility. We are inevitably led to conceive of an object which appears but we cannot admit its actual existence as quite independent of our sensibility. Such things in themselves, we might say, are ‘necessarily conceivable’ but ‘in principle unknowable’; they remain for us in a condition of ‘necessary conceivability and principled ignorance’ (562–4). These background temptations and pressures are captured in my references to the ‘bewitchment of intelligence by language’ (590, 596), in which we may be philosophically misled by this powerful but misleading ‘conceptual shadow’ (256–8, 400–2, 451). Part of that characterization, necessary conceivability, explicitly allows a more positive use for the philosophical conception of things in themselves (signalled at Bxxvii–xxx), but it raises the question whether in that more positive use Kant violates the limitations implicit in the other part, namely principled ignorance. That is the issue in 3.2.
Understood in this way is the absurdity argument ‘very weak’? There is no doubt that it then forms a vital part of Kant's basic critique of reason and of earlier metaphysics. More than that, however, the objection that its weakness rests on its ‘grammatical’ character can be answered in two ways. First, to repeat, the point can be put just as well without relying specifically on the purely grammatical aspect of the term ‘appearance’. Second, the term ‘appearance’ is an entirely natural, even inevitable, way of identifying what belongs distinctively to the senses. It is a term which philosophers generally, in Kant's time and ours, might readily accept as standing for a specifically sensory contribution to experience. What is distinctive in Kant is not his choice of such a term but the use he then makes of it, both to contrast an inarticulate, indeterminate, sense and a differentiating understanding, and to separate the colloquial and philosophical contexts of its use (B59–66, B69–71).
3.2 Why we might Endorse a More Positive Use for ‘Things in Themselves’
I have already made clear that I agree with Manfred Kuehn in his view that Kant speaks of ‘two worlds’, and allows a more ‘positive’ appeal to things in themselves beyond its negative and therapeutic value. I take the problem to be only what kind of commitment is additionally made in such a use. In TRK I noted the wide range of such appeals in, for example, the notion of a metaphysical Grund for experience, the Ideas of systematic unity, the highest good, a divine personality and governing ideals in psychology, cosmology and elsewhere. But the most evident case for Manfred Kuehn, which Kant highlights immediately after the ‘absurdity’ argument (Bxxvii–xxviii), is that of morality. In the Third Antinomy, in the Grundlegung and second Critique, Kant makes that case and plainly uses the concept of an intelligible world, a kingdom of ends, with its complement of things in themselves. The complaint is that I did not do justice to that more positive appeal and so somewhat distorted it. Though I recognized additional pressure from the side of morality to confirm the conclusion of the ‘speculative’ resolution of the Third Antinomy, I did not pursue that moral issue (689).
Here, too, I confine myself to answering, in schematic form, two questions. (a) Does Kant retract or violate the earlier restrictions placed on any positive use of ‘things in themselves’? (b) How does the more positive use of ‘thing in itself’ function in the account of freedom and morality? The answers that I offer are those already given in TRK. My belief is that, despite the common view of Kant as in this context entangled in confusion and incoherence, he does not, and does not need to, retract the restrictions or violate them. I also believe, though I am less sure of this, that Manfred Kuehn might agree with me about this.
(a) That Kant does not retract his earlier restrictions on our knowledge of things in themselves is shown in the following passages: In the Grundlegung (4: 451): ‘[E]ven with the most strenuous attention … we can achieve cognition only of appearances, never of things in themselves.’ In the passage Kant repeats his earlier view, in the grammatical or non-grammatical form, of the necessity of conceiving things in themselves: ‘[I]t follows that we must assume behind appearances something else that is not appearance, namely things in themselves, although … we resign ourselves to being unable to come any closer to them or … to know what they are in themselves.’ The mere ‘thinkability’ of things in themselves, however necessary, is underlined at 4: 458 and 5: 55: ‘By thinking itself into a world of understanding practical reason does not at all overstep its boundaries … causa noumenon … is, though a possible thinkable concept nevertheless an empty one.’ In the Grundlegung (4: 462) the restrictions are forcefully and vividly repeated in an echo of the anti-Platonic ‘dove’ imagery of B8–9:
… even though I have an Idea of it [the intelligible world], yet I have not the least cognizance of it … Here … is the highest limit of all moral enquiry and it is already of great importance to determine it so that reason may not, on one hand, search about in the world of sense for the supreme motive and a comprehensible but empirical interest, and may not, on the other hand, impotently flap its wings without moving from the spot in the space, which is empty, of transcendent concepts called the intelligible world, and so lose itself among phantoms.
Kant there underlines the defects both of merely empiricist and of pseudo-rationalist, transcendent accounts of morality. It is not surprising, then, to find him repeating the same restrictions in the second Critique 5: 57:
… as for whatever other properties, belonging to the theoretical way of representing such supersensible beings, may be brought forward … these are without exception to be counted not as knowledge but only as a warrant (for practical purposes a necessity) to admit and presuppose them … so … pure theoretical reason is not given the least encouragement to fly into the transcendent.
(b) It may be suggested that the restrictions imposed on a theoretical use of reason in this way nevertheless give way, according to Kant, to a practical knowledge of freedom and moral agency in which we ascribe moral features to things in themselves in an intelligible world. In the passage just quoted Kant talks only of a ‘warrant’ in presupposing an application to the super-sensible world, but this raises a question about the character of such a warrant. I claim that Kant is not committed to endorsing moral judgements about an intelligible world, and outline the structure of the argument from the resolution of the Third Antinomy to the applied moral philosophy.
What Kant claims to have established in his resolution of the Third Antinomy is, as is by now well known, severely limited. At B585–6 he denies even aiming to establish the ‘reality’ or the ‘possibility’ of (transcendental) freedom, and these restrictions are repeated in the later moral philosophy, for example, Grundlegung 4: 455: ‘Freedom is only an Idea whose objective reality is doubtful.’ What the arguments are designed to establish is only the ‘compatibility of freedom and nature’ (B586) in a position Kant importantly calls ‘defensive’ (4: 459). They show, as he says (5: 93–4), that ‘no insight can be had into the possibility of freedom as an efficient cause’ but only a ‘sufficient assurance that there can be no proof of its impossibility’. The point is repeated in the Grundlegung 4: 459:
Freedom … is a mere Idea the objective reality of which can in no way be presented in accordance with the laws of nature … Now where determination by laws of nature ceases there all explanation ceases as well, and nothing is left but defense, that is, to repel the objections of those who pretend to have seen deeper into the essence of things and therefore boldly declare that freedom is impossible.
That limited, defensive, strategy demonstrates the boundary between what is for us knowable and what is not (4: 458, 462). It entitles us to conceive of the ascription of both natural causality and transcendental freedom to one and the same being, but it does not explain any further the link with our actual moral responses and judgements. What is additionally needed is an argument from morality to give substance to, or ground for, our ‘interest’ (4: 459) in realizing the Idea of (transcendental) freedom in morality and the law (5: 93–4, 43). The point is clarified in the second Critique (5: 105) where the ‘principle of morality’ provides that transition from a merely ‘thinkable possibility’ (‘can’) to an actuality (‘is’):
Therefore that unconditioned causality and the capacity for it, freedom, and with it a being (I myself) that belongs to the sensible and to the intelligible world … is not merely thought indeterminately but is determined with respect to the law of its causality and cognized assertorically … which for theoretical purposes would be transcendent [but] is for practical purposes immanent.
Such a development in the realization of transcendental freedom still does not entitle us to claim to know the intelligible world theoretically. In particular, I believe, it does not entitle us to apply our moral judgements to such a world rather than to the world of sense viewed under the assumption of transcendental freedom. The point can be made by comparing passages from William James and Kant. James, in what he calls his ‘crass supernaturalism’, gave an account of a spiritual, religious world in the following terms: ‘The world interpreted religiously is not the material world over again with an altered expression … it must have a natural constitution different from that which a materialist world would have’ (James Reference James1902: 518).
The suggestion is that such a spiritual world has to be accepted as distinct from the material world of the senses in a literal ‘two world’ view. Kant, by contrast, seems to express a different, and more integrated, relation between the world of the senses and our moral judgements at B835–6 and in the second Critique (5: 43). There he reiterates the idea that the ground for an ‘intelligible’ world rests on the argument from morality, and goes on to characterize the upshot in our ascription of morality. He says:
The Idea of a moral world has, therefore, objective reality, not as referring to an object of an intelligible intuition (for we are quite unable to think any such object), but as referring to the sensible world viewed as being an object of pure reason in its practical employment. (B836)
This law is to furnish the sensible world, as a sensible nature (in what concerns rational beings), with the form of a world of understanding, that is of a supersensible nature, though without infringing on the mechanism of the former. (5: 43)
In Kant's account the upshot of the defence of freedom is only to integrate our moral, intelligible and our sensory, material judgements; it is not to divide or separate the two realms. Those two ways (standpoints) of talking (conceiving) are required to resolve the Antinomy, and minimally to defend a transcendental, super-sensible freedom, but it is recognition of the moral law which then allows us to talk immanently, as we do, of a human sensible nature subject to the demands of morality. Kant's arguments in his defence of transcendental freedom may not work, and his appeal to the fact of moral law may itself be flawed. But the consequence, in terms of his commitment to an intelligible world, is not very different from that in the theoretical context in which unschematized categories call on an intermediary schematism in order to realize them in a usable form in immanent experience. The difference is that we have an ‘interest’ in appealing to a super-sensible freedom, as a defence against determinism in the Antinomy, while we have no theoretical interest in appealing to a super-sensible world of unschematized categories.
You have
Access
I am genuinely grateful to the critics both for their compliments, their careful attention to TRK, and for their criticisms. The latter give me an opportunity to clarify some misunderstandings and, where agreement seems impossible, to explain again my position and the grounds for it. When I first read the comments I did not notice any particular theme connecting the contributions, but now it seems to me that there is one. It starts with ambiguities in the Second Analogy's principle, some of which link directly to determinism and the resolution of the Third Antinomy, goes on to issues about apriority and the significance of Kant's distinction between appearances and things in themselves, and finally issues in different views about the ‘unconditioned’ and our knowledge of, or belief in, such an item especially in its moral context. Though there is in the contributions only a muted hostility to my general distinction between a ‘traditionalist’ and a ‘revolutionary’ approach to Kant, this persistent theme allows me to clarify one of its central aspects in the transition from speculative to practical philosophy and the resulting commitment to things in themselves.
1. Response to Gary Banham
Gary Banham has correctly characterized the central aspects of my account of the Second Analogy as ‘minimalist’, ‘austere’, ‘abstract’ and ‘formal’ (but not sceptical as p. 204, above, suggests: ‘we will, necessarily, never have assurance when we are actually experiencing an event’). I do not believe that the principle of the Second Analogy (the ‘general causal principle’, GCP) entails any principle of the uniformity of nature or provides any guarantee of the future truth of specific causal laws against a Humean scepticism about induction. The primary difference between Kant and Hume, as I represent it, lies in Kant's applying the notion of ‘necessity’ initially to the a priori GCP and then only derivatively to specific a posteriori causal laws. Hume by contrast is principally concerned with the claim, which he rejects, that necessity attaches to specific causal laws, and with a primacy accorded to those specific laws over the GCP. Nor do I identify the understanding's constitutive GCP with any regulative principle of reason such as that of universal determinism in the Third Antinomy. In my minimalist account the GCP is claimed to express a non-analytic but necessary connection between the concepts of a cause and an event. Gary Banham is quite right to describe that connection, as I represent it, as one not between events but within an event between different states of some substance.
At almost all these points there is a tradition which takes a contrary view and Gary Banham includes himself in that tradition. In the tradition it is often recognized that the GCP does not entail any principle of the uniformity of nature which might answer Hume's inductive scepticism, but this is regarded as an objection to Kant's supposed attempt to provide such an answer. Since I deny that the GCP is directed at such an anti-sceptical response to Hume I reject this standard objection to Kant's position. Gary Banham's position is that the GCP must offer an answer to inductive scepticism, that it must be strengthened in order to rebut Hume's denial of a necessity in specific causal laws and that it entails a universal determinism considered in the Third Antinomy. In short his criticism is generally to the effect that my understanding of the GCP and its role in Kant's argument is too weak. There are some misunderstandings in his arguments against my position, which I try to correct first, but there are evidently also central disagreements between us which I then consider.
1.1 Some Misunderstandings
Gary Banham rests a number of objections on his appeal to types and tokens. One suggestion is that although I make a passing reference to types (457) I do not make any references to specific event-types and so disregard the connection between Kant's causal principle and specific causal laws. But there are also suggestions that my account of Kant's causal principle concerns only token events and so cannot be generalized or support a principle of universal determinism. But for me Kant's principle is a general claim about ‘objects’ of the type ‘event’. What it expresses holds for every event in virtue of a synthetic necessary conceptual connection between ‘event’ and ‘cause’, in which the concept of a cause is required to make possible objects of the type ‘event’. The principle does not, except illustratively, appeal to any specific causal laws, or indeed to any specific event-types, for such appeals would require a posteriori material. Kant makes this point in the Principles (e.g. B252: ‘How anything can be altered … we have a priori not the least conception … [without] knowledge of actual forces which can be given only empirically’) and in the Preface of the Metaphysical Foundations of Natural Science. In the formal proof references to ordered states of some substance in an event are characterized only by variables (‘A’ and ‘B’ (B237); ‘a’ and ‘b’ (B253)). The synthetic necessary conceptual connection on which the principle rests is in that way quite general, but it does not entail a universal determinism as I claim below.
1.2 The Principle of Continuity (B253–4)
Gary Banham raises other questions about the GCP's relation to generalizations such as a principle of universal determinism or a principle of the systematic unity of nature (discussed respectively in the Third Antinomy and the Appendix to the Dialectic). My view is that the GCP does not entail universal determinism and that the GCP differs from a principle of systematic unity in being constitutive rather than regulative (see section 1.3). To support the link with determinism he cites a passage from B253–4 about a principle of continuity expressed as: ‘All alteration is … only possible through a continuous action of causality.’ Two questions arise here. (i) Does the passage, with its reference to a principle of continuity, carry such a commitment?, and (ii) Does the GCP anyway express or entail such a universal causal determinism? In both cases my answer is negative.
1.2.1
The quotation comes in a passage which takes its starting point from a question about the transition from one state a to another b in a causal relation. The question plainly refers to the central issue in Kant's argument, namely the requirements which have to hold for a determinate, objective, time order to hold between two states in some substance. Kant is, in the passage, primarily concerned to claim that such a transition occurs over time, that it is not sudden or instantaneous, but goes through an infinite set of times between the two limits to the event, namely the starting point a and the terminal point b. Those claims make a reference both to the mathematical principles, especially the Anticipations, and to the application of the calculus to such causal relations. In the Anticipations (B211–12) Kant insists that there are no smallest possible times, and that spatial points and temporal instants are only ‘limits’ to some continuing, extensive, phenomenon, like that between the states a and b in some substance. This link with the earlier mathematical principles is valuable because while commentators nowadays commonly, and rightly, insist on the inter-relationship between all three Analogies they do not always note the equally strong inter-relationship between Analogies and Anticipations. Understood in this way the principle of continuity applies in the passage to ‘micro-events’, that is, to the nature of a causal transition within events, rather than to a global universal determinism across events.
1.2.2
There is, in any case, no inference from a principle of continuity, so understood, to such a universal determinism. It may be that within any such causal transition the continuity principle ‘determines’ that outcome, but this is to say no more than Kant is already committed to, namely that the transition is governed by causal law. It does not follow that every single state of the universe is determinable or predictable from an earlier state and a set of causal laws. Nor is there any such entailment from the GCP, as I have understood it, to that universal determinism. The GCP expresses only a synthetic necessary connection between the concepts ‘cause’ and ‘event’ that might hold even if there were gaps in the universe's causal record, and even if some such transitions were statistical rather than causal.
It may be thought perverse to deny any connection between GCP and determinism but my account does not go that far. My view is that the GCP naturally allows, and even encourages, us to conceive of such a determinism, which then potentially conflicts with the ascription of freedom and agency (see the response to Manfred Kuehn). But on my account such a determinism is not a constitutive truth which we might be able to establish, but a regulative principle which enjoins us to extend the search for causes as far as possible (as in the cosmological Antinomy). Gary Banham himself seems to express the relation, between the GCP and determinism, in terms weaker than entailment which conform to my view. He speaks of the Analogies ‘com[ing] together in a “picture” of a dynamical world-whole’, of linking ‘the three Analogies to the thought’ of such a whole, and of their ‘lead[ing] naturally to a conception of universal determinism’. At this point there is a welcome convergence between our views.
1.3 Constitutive and Regulative Principles and Kant's Relation to Hume
Just as Gary Banham criticizes my belief that the GCP does not entail universal determinism, so he also objects to my view that the principle of systematic unity, as a principle of reason rather than of understanding, is not, and is not intended as, an answer to a Humean inductive scepticism. The two complaints are evidently related since the GCP, for example in its SCSE (‘same cause same effect’) form, may also be taken as a response to Hume's inductive scepticism. That objection rests mainly on the claim that ‘it is not on the grounds of reference to [constitutive principles of understanding] that we can establish a view of empirical truth’, but rather by reference to reason's principle of systematic unity (B679) (p. 206). There is one unintended fault here which deserves to be quickly corrected and forgotten, namely in the suggestion that the constitutive principles of understanding do not provide any ground for empirical truth. That is not what the passage cited from B679 says, but it is surely evident that Kant regards principles of understanding as providing a general structure which grounds empirical truth. The passage at B247 specifically appealing to the GCP makes this clear: ‘[T]he relation of cause to effect … is the condition of the objective validity of our empirical judgments … and so of their empirical truth.’
Evidently what is at issue is not whether those constitutive principles relate to empirical truth but whether in addition the regulative principles of reason also provide some such ground. Gary Banham is entirely right to refer to the passage at B679 to indicate such a view (he might also have cited B681, B685). In those passages Kant offers a form of ‘transcendental argument’ to show that the principle of systematic unity provides a condition for empirical truth, and I offered a detailed account of them elsewhere (1999: 31–45). Those Kantian arguments do not conflict with the account of transcendental principles of the understanding, but Kant makes it clear that the two kinds of principle, constitutive principles of the understanding and regulative principles of reason, have different roles and outcomes.
In particular, in passages to which Gary Banham makes no reference, Kant makes clear that the regulative principles are not themselves objective, constitutive truths and are not susceptible of a strict transcendental deduction along the lines of the principles of the understanding (B691–2). The ‘transcendental arguments’ of B679, 681, 685 may have the same form as that of a transcendental deduction but they do not ascribe the same character or status to the two types of principle. The regulative principles, called ‘maxims of reason’, are described as ‘subjective principles derived not from the constitution of an object but from an interest of reason’ (B694). They possess ‘objective but indeterminate validity’ (B691) and gesture towards an imaginary end-point, a ‘focus imaginarius’ (B672). They apply to objects only through their heuristic influence on the understanding and its principles. Kant's final answer to his central question in the Dialectic – ‘Is reason an independent source of concepts … [which] relate to objects?’ (B362) – is an unequivocal ‘Reason is never in immediate relation to an object but only to the understanding … it does not create concepts but only orders them’ (B671). Other such provisos are given at B693, B697, B675, B702.
Kant's position is that of a philosopher who acknowledges that we can have no access to a final, complete, truth even in science, but who nevertheless insists that it is incoherent to carry out scientific enquiry without being motivated by a search for such an ideal. Such a conception of a completed truth functions, however, only as a heuristic ‘focus imaginarius’. We are guided by such a motive, with its ‘indeterminate validity’, but have no assurance of the truth, or proof, of its related principle. It represents, properly, a maxim of reason reflecting an ‘interest’ rather than a categorial objective principle, and appeals not to a genuine schema in relation to sensibility but only to the ‘analogon of a schema’ (B693). These extensive qualifications make it impossible to regard a maxim of systematic unity as a constitutive truth or as a guarantee of scientific, inductive, success in answer to Hume's scepticism.
A similar neglect of a key passage arises from Gary Banham's account of Kant's disagreement with Hume. Gary Banham argues against my view that the basic necessity belongs to the GCP and then only derivatively to specific causal laws. He thinks that I disagree with the view quoted from Michael Friedman (p. 207–208), but that is not so. I agree with the view expressed in the quotation which takes the necessity of the GCP as fundamental and that of specific laws as derivative. The derivation is straightforward in that specific laws are in principle deducible from the GCP together with relevant a posteriori material. It is in that sense, and with that dependent relationship, that Kant can speak, as he does, of a ‘dignity’ attaching to specific causal laws which cannot be empirically expressed, in virtue of their dependence on the GCP (B124). The dignity, necessity, of the GCP trickles down, as it were, to the character of the specific laws falling under it. The alternative position, which Gary Banham endorses, is to ascribe necessity without qualification and underivatively to specific causal laws, although both Kant and Hume regard these as a posteriori.
That alternative is, I believe, incompatible with the passage at B793-4 where Kant looks back at his disagreements with Hume and makes it plain that over ‘cause’ the disagreement arises from the fundamental ascription of necessity to the GCP and not to specific causal laws, whose dignity, and dependent ‘necessity’, is only derivative. Hume plainly has no adequate apparatus available to recognize Kant's point. For Hume specific causal laws, and the GCP, must be, and be derived from, either analytic (a priori) or synthetic (a posteriori) resources. Kant's non-empiricist conception of the synthetic a priori enables him to classify, and understand, both the GCP and specific causal laws, and their relationship, in a more sophisticated way. The basic disagreement between Kant and Hume over cause lies evidently in those alternative, incompatible ways of ascribing a fundamental or dependent necessity either to the GCP or to specific laws falling under it.
2. A Response to Gordon Brittan
I am grateful for Gordon Brittan's complimentary remarks and for his choosing to discuss what I called Kant's ‘analytic apparatus’. I found particularly valuable his critical comments on the comparison between Kant and Kripke. With almost all of what he says I find it easy to agree, not least because his summary of my views is, with one possible exception, both clear and accurate. His principal criticisms are: (1) my account of a priori truths (judgements) is inadequate, and confuses what might be called a ‘formal’ and a ‘functional’ treatment of the synthetic a priori classification; (2) I fail to note a connection between analytic truth and existence and so overlook part of the comparison between Kant's account of singular reference and Kripke's account of rigid designation.
2.1 Formal and Functional Accounts of the Synthetic A Priori Classification
Gordon Brittan notes two different aspects of my account of the synthetic a priori classification; one focuses on the formal criteria for the two separate classifications (synthetic/analytic and a priori/a posteriori), and the other presents an informal and functional picture of the way in which the resulting hybrid classification (synthetic a priori) figures crucially in Kant's understanding of ‘conditions of possible experience’. Gordon Brittan rightly stresses the central location reserved in the Critique for the question: ‘How are synthetic a priori judgements possible?’ He notes also how, in my discussion, I offer an initial formal characterization of the hybrid (in comments on the Introduction) and reserve for later (especially in comments on the Analytic of Principles) discussion of the way that characterization works out in the crucial cases. He notes, too, quite correctly, that even in the former comments I refer to a familiar, Wittgensteinian, informal and functional analogy with the rules of a game such as chess. But he thinks that this is a confusion which is part of a multiple ambiguity in Kant's account of the a priori (p. 211, p. 215).
I think that this objection relates to one important misunderstanding in confusing ‘synthetic a priori’ with ‘transcendental’. Kant's view, I believe, is that the ‘transcendental’ principles, identified and proved in the Analytic, are ‘synthetic a priori’, but that these two characterizations are not equivalent. Transcendental principles, which make experience possible, are synthetic a priori but not all synthetic a priori judgements are transcendental or make experience possible. The point is made at B80–1 and discussed in TRK, chapter 5. Many synthetic a priori judgements, like most of those asserted in mathematics or physics, are not transcendental, and do not have the role of making experience possible. The ‘formal’ characterization attaches properly to ‘synthetic a priori’ and the ‘functional’, Wittgensteinian, account to what is ‘transcendental’.
For this reason the extensive ambiguity in ‘a priori’ which Gordon Brittan marks (p. 218) is not as extensive as he claims. One such ambiguity is resolved by distinguishing the two non-equivalent classifications ‘synthetic a priori’ and ‘transcendental’. But even the other listed ambiguities are resolvable; for the diverse references to types of a priori all seem to me to come under the general characterization of ‘independence of experience’, applied in diverse contexts (cf. TRK, ch. 4, 65). That is true of an a priori intuition's non-causal origin and of non-transcendental but a priori judgements whose warrant does not rest on what I called ‘presented experience’.
Gordon Brittan disapproves of that epistemic characterization because the notion of ‘presented experience’ is opaque and ripe for philosophical analysis even on Kant's terms. Moreover it does not by itself indicate a connection with universality and necessity, the crucial marks of the a priori. These anxieties are somewhat lessened by recognizing the distinction between the ‘synthetic a priori’ and the ‘transcendental’. Kant's distinctively metaphysical position focuses on those elements which are transcendental conditions of experience, and on their allocation to different faculties of sensibility, understanding, and reason. When those elements are expressed in judgemental form they are then claimed to be synthetic a priori, and can be tested for that character. Moreover Kant's understanding of what is presented in experience makes a minimal link with necessity and universality in virtue of his belief that such presentations cannot support a strict necessity or universality. What is then an uncontested claim to apriority, as in the illustrative accounts of mathematics, allows such a strict necessity or universality. It is not that in such an account merely apparent universality and necessity are taken as decisive criteria for apriority, as in Bennett's criticism of Kant's alleged ‘introspective semantics’ (Bennett Reference Bennett1966, 7–8). Nor is it the case that classifying a judgement as ‘a priori’ already establishes its truth, since we can know judgements to be false a priori. The a priori classification by itself is not an anti-sceptical guarantee of truth. Establishing the truth of an a priori judgement will depend on a proof providing the right kind of warrant.
2.2 Analytic/Synthetic Truth and Singular Reference
Gordon Brittan has several understandable qualms about the way I characterize the analytic/synthetic distinction (on Kant's behalf), and these lead on to his central complaint about a failure to recognize significant points about singular reference. When he says (p. 214) that my account ‘deserve[s] more discussion’ I entirely agree. My aim was consciously limited to giving an account of Kant's position, and defending it against some standard objections, rather than to offer an account which might satisfy contemporary logicians. My principal response is to acknowledge the value of an appeal to aspects which I neglected but to enter one caveat about their relation to Kant's discussion. I am not entirely persuaded that Kant's position is just as Gordon Brittan claims.
The central objections raised in this context go back to reservations about the ways I characterize the analytic/synthetic distinction. I distinguished different, but equivalent, formulations of the distinction according to their application to judgements or to truth and falsity. One reason for this was to avoid a common error in identifying analytic judgements as those whose denial yields a contradiction, which entails absurdly that analytically false judgements are synthetic. Gordon Brittan's objection begins with a difficulty in appealing to the truth (or falsity) of analytic judgements. He appeals to Kant's recognition of a ‘nominal’ definition of ‘truth’ (B82) as ‘agreement of cognition with its object’ in order to rest ascription of analyticity on the supposition that the terms in the relevant judgement genuinely refer. He, therefore, talks sometimes only of the ‘validity’ of analytic judgements, though he also recognizes their ‘necessary truth’. The latter, he says consequently, has to be expressed in the claim ‘If a judgement's terms refer, then its truth can be determined via the law of contradiction.’ Except for the addition of the antecedent this formulation is virtually the same as one I offered on Kant's behalf in terms of true analytic judgements (TRK, 81).
The reason for adding that antecedent rests initially on a passage quoted from B314 in which Kant says that an analytic judgement ‘leaves it undecided whether the [subject] concept even has any relation to objects … [W]hat the [subject] concept might pertain to is indifferent’. But this claim can equally naturally be read as claiming that the reference of the subject term is irrelevant to the determination of the truth (falsity) of an analytic judgement. In such cases (‘The unicorn has one horn’) the predicate can be (metaphorically) said to be ‘contained in’ the subject term, or its denial can be shown to yield a contradiction, even if there are no unicorns.
Gordon Brittan appeals also to a further claim about the ‘knowability’, or ‘use’, of analytic judgements, and cites a passage at B15 about the ‘use’ of such judgements in mathematical proofs. On my account there is no obstacle to saying that we can know the truth (falsity) of an analytic judgement even when its subject term does not refer. The mathematical case does not deny the knowability of such judgements so much as their use in establishing further mathematical truths. Gordon Brittan is understandably influenced by the Kantian claim that we can have no knowledge of super-sensible objects, but there is no necessity to carry this over to such analytic judgements. In the theological context (which he also cites) can Kant not say that analytic truths about God can be known even though that does not entitle us to claim to know of His existence? If that option remains available, then Gordon Brittan's alternative is at least not obligatory.
One important corollary of Gordon Brittan's position is, as he says, that for Kant ‘no analysis of concepts determines reference’, and he links this to Kripke's non-descriptive account of names and to a contrary Leibnizean view. The point is important because to interpret Kant in this way is to offer an argument for an essential appeal to intuition over and above concepts in meeting the requirements for genuine experience. I have to admit that in my comparison of Kant and Kripke I had not made the point in this way. The closest I got to it was to mention Leibniz's conception of a complete concept for each individual as an essential target in Kant's criticisms of Leibniz in the Amphiboly (837 n. 18 and 543–7). Despite this omission, however, I think that the argument and consequence which Gordon Brittan rightly notes are just as evident in other parts of Kant's discussion, such as his contrasts between the roles of concepts and intuitions (B74–9, B91–4), between treating logic as a mere ‘canon’ or as an ‘organon’ (B87–8), and in his contrast between logical and transcendental reflection in the Amphiboly (B324–36).
The first point offers an account of concepts as ‘mediate’ ‘representations of representations’ which denies them the function of directly referring to objects, and underlines the familiar need for both concepts and intuitions in experience to avoid ‘blindness’ and ‘emptiness’ (B75–6). The second insists on the limitations of a purely logical account of concepts, and on the dangers of inferring directly from concepts and their analysis to truths about objects. That point is presented by Kant as the basis for dialectical illusion (see my response to Adrian Moore). The third is the ground for Kant's belief that Leibniz's metaphysical errors all stem from his failure to recognize a distinction between logical reflection which disregards reference to objects and transcendental reflection which insists on such an appeal.
A consequent objection is that in claiming that Kripke's notion of rigid designation ‘plays no part in Kant’ I failed to exploit this aspect of the comparison. Gordon Brittan agrees with me (p. 219, above) that Kant's ‘synthetic a priori’ and especially ‘transcendental’ principles can be understood as holding not in all possible worlds but only in some subset of them defined in principle by reference to our distinctively human cognitive capacities. But he thinks that Kant opposes both Kripke and Leibniz in denying that ‘there are any genuine singular terms’ which pick out the same individual in all possible worlds. I have two related difficulties with this. First while I agree that Kant effectively deploys the notion of judgements holding in some or all possible worlds I do not see any of the machinery required specifically to differentiate reference in some, or all, possible worlds. Kant, as I claimed, does not explicitly distinguish different kinds of referring expression, and he has no indication of issues arising from trans-world identity. For these reasons I would be inclined to put this aspect of Kant-interpretation into the category of what I called ‘developments of Kant's insights’ (TRK, 1) rather than a direct expression of what he either actually says or means.
Second, and more generally, although Gordon Brittan's interpretation is genuinely illuminating it seems to me to put too much weight on the linguistic version of Kant's account, as does his overall characterization of the Critique's project as a ‘theory of objective reference’. Gordon Brittan admits that such an account ‘rephrases’ that project ‘in a more linguistic mode’, which is naturally congenial to our contemporaries and may have real value (as it does in this case). But it is liable to distort Kant's own position for it is not an oversight, but fundamental, for Kant to allocate transcendental elements of experience to one or other psychological faculty. Commentators avoid that psychological mode because it is believed to incorporate the errors of ‘psychologism’. But in TRK I claimed not only that these errors can be avoided but also that Kant's psychological formulation acts usefully as a counter-weight to a twentieth-century over-emphasis on language. There is, I believe, a real point in retaining that Kantian psychological mode.
3. Response to Manfred Kuehn
There is little in Manfred Kuehn's account that I disagree with, and despite some misunderstandings I believe that there is a strong convergence in our positions. Manfred Kuehn's title indicates two different, but related, aspects of Kant's account of appearances and things in themselves. There are questions about ‘how’ we might come to think of things in themselves, and questions about ‘why’ we might endorse, in some way, an appeal to such things. In the former Manfred Kuehn thinks one answer, from the ‘absurdity’ argument (Bxxvi–xxvii), is demonstrably weak; in the latter he thinks that I do not do justice to Kant's commitment to a world of things in themselves, especially in the moral philosophy. These are the two topics I consider.
3.1 Appearances and Things in Themselves: How we Come to Think of the Latter
There are some misunderstandings here. Manfred Kuehn asks whether my criticism of the conventional ‘two worlds/two aspects’ labels is directed against Kant or against his commentators. I had hoped to make plain (e.g. 550) that in my view the misuse of that contrast marks errors among commentators rather than in Kant. That misuse has at least two aspects: that the contrast is used, first, without proper clarification of its sense, and, second, with the (mistaken) tacit assumption that we have to choose between the two options. In the first it may not be made clear what the commitments of the alternative views are. If we ascribe a ‘two worlds’ view to Kant does this commit him to the actual existence of things in themselves or only to their conceivability? Should it not be made clear that even if Kant accepts only the conceivability of ‘two worlds’ it is nevertheless two worlds that are conceivable? In the second it may not be recognized that, as Manfred Kuehn and I both point out, Kant demonstrably uses both conceptions. These are standard examples of a careless appeal to conventional labels but they are enough to show how misleading and erroneous such appeals can be. My target is that reference to what have sometimes become, in the literature, thoughtless labels; and it is those thoughtless labels that should be dropped.
My account of the ‘absurdity’ argument provided the opportunity to underline those errors. In Kant it provides an account not only of how we do come to think of things in themselves but also of how we must come to think of them. Formally the argument does not contain any further indication of more positive uses of that necessary conception although Manfred Kuehn is quite right to insist that such positive uses are indicated in what follows (Bxxvii–Bxxx). Manfred Kuehn regards the absurdity argument as ‘very weak’ primarily because it rests on a ‘grammatical’ point about the terms ‘appearance/appears’, and then the question arises: why should we use the term ‘appearance’ at all? He is right that the central point is ‘grammatical’, but I think his objections can be answered.
The first thing is to insist that the argument does not entail commitment to the actual existence of things in themselves (and I think Manfred Kuehn agrees). The second point is to relate the (grammatical) argument to Kant's underlying distinction between sensibility and understanding. Although the point is put grammatically in the B Preface it can be made just as well, as it had been made in A, through the distinction between sensibility and understanding. For Kant ‘appearances’ are supposed, strictly, to belong to sensibility without any reference to the descriptive (conceptual) resources of the understanding. This requirement, artificial and theoretical as it may seem, is captured in his account of appearances as ‘indeterminate’ objects of intuition (B34), which points inevitably to the need to explain that further differentiation into objects which appearances are appearances ‘of’. The term Kant uses to mark the need, in this passage and elsewhere, is ‘thing in general’ (Bxxvii), while ‘transcendental object = X’ has a similar role in the Deduction (A104–6, A109). Kant's subsequent point is that these pressures, whether grammatical or not, can lead philosophers in two radically different directions. In one it leads to a conception of objects wholly distinct from, independent of, our senses and accessed via concepts alone through understanding or reason; in the other it leads to a conception of objects accessed by sensibility and differentiated, on that assumption, through the conceptual resources of understanding or language. The grammatical pressures in the argument acknowledge both the necessity of an appeal to that differentiation into objects, but also the potential error of taking such objects as wholly independent of the senses.
Such an account trades on Kant's later treatment of this underlying philosophical error in the Dialectic and in the Logic (B86–8, B186–7, B342–6). It trades also on his discussion of the Aesthetic in the General Observation where Kant distinguishes the colloquial and the philosophical contexts of talk about things and the way they appear (B62–3, B69–70). In those passages the lesson is that whereas colloquially we are entitled to admit the actual existence of things that appear to us (empirical objects of experience), in the philosophical context we cannot do so if those things are taken to be independent of sensibility. We are inevitably led to conceive of an object which appears but we cannot admit its actual existence as quite independent of our sensibility. Such things in themselves, we might say, are ‘necessarily conceivable’ but ‘in principle unknowable’; they remain for us in a condition of ‘necessary conceivability and principled ignorance’ (562–4). These background temptations and pressures are captured in my references to the ‘bewitchment of intelligence by language’ (590, 596), in which we may be philosophically misled by this powerful but misleading ‘conceptual shadow’ (256–8, 400–2, 451). Part of that characterization, necessary conceivability, explicitly allows a more positive use for the philosophical conception of things in themselves (signalled at Bxxvii–xxx), but it raises the question whether in that more positive use Kant violates the limitations implicit in the other part, namely principled ignorance. That is the issue in 3.2.
Understood in this way is the absurdity argument ‘very weak’? There is no doubt that it then forms a vital part of Kant's basic critique of reason and of earlier metaphysics. More than that, however, the objection that its weakness rests on its ‘grammatical’ character can be answered in two ways. First, to repeat, the point can be put just as well without relying specifically on the purely grammatical aspect of the term ‘appearance’. Second, the term ‘appearance’ is an entirely natural, even inevitable, way of identifying what belongs distinctively to the senses. It is a term which philosophers generally, in Kant's time and ours, might readily accept as standing for a specifically sensory contribution to experience. What is distinctive in Kant is not his choice of such a term but the use he then makes of it, both to contrast an inarticulate, indeterminate, sense and a differentiating understanding, and to separate the colloquial and philosophical contexts of its use (B59–66, B69–71).
3.2 Why we might Endorse a More Positive Use for ‘Things in Themselves’
I have already made clear that I agree with Manfred Kuehn in his view that Kant speaks of ‘two worlds’, and allows a more ‘positive’ appeal to things in themselves beyond its negative and therapeutic value. I take the problem to be only what kind of commitment is additionally made in such a use. In TRK I noted the wide range of such appeals in, for example, the notion of a metaphysical Grund for experience, the Ideas of systematic unity, the highest good, a divine personality and governing ideals in psychology, cosmology and elsewhere. But the most evident case for Manfred Kuehn, which Kant highlights immediately after the ‘absurdity’ argument (Bxxvii–xxviii), is that of morality. In the Third Antinomy, in the Grundlegung and second Critique, Kant makes that case and plainly uses the concept of an intelligible world, a kingdom of ends, with its complement of things in themselves. The complaint is that I did not do justice to that more positive appeal and so somewhat distorted it. Though I recognized additional pressure from the side of morality to confirm the conclusion of the ‘speculative’ resolution of the Third Antinomy, I did not pursue that moral issue (689).
Here, too, I confine myself to answering, in schematic form, two questions. (a) Does Kant retract or violate the earlier restrictions placed on any positive use of ‘things in themselves’? (b) How does the more positive use of ‘thing in itself’ function in the account of freedom and morality? The answers that I offer are those already given in TRK. My belief is that, despite the common view of Kant as in this context entangled in confusion and incoherence, he does not, and does not need to, retract the restrictions or violate them. I also believe, though I am less sure of this, that Manfred Kuehn might agree with me about this.
(a) That Kant does not retract his earlier restrictions on our knowledge of things in themselves is shown in the following passages: In the Grundlegung (4: 451): ‘[E]ven with the most strenuous attention … we can achieve cognition only of appearances, never of things in themselves.’ In the passage Kant repeats his earlier view, in the grammatical or non-grammatical form, of the necessity of conceiving things in themselves: ‘[I]t follows that we must assume behind appearances something else that is not appearance, namely things in themselves, although … we resign ourselves to being unable to come any closer to them or … to know what they are in themselves.’ The mere ‘thinkability’ of things in themselves, however necessary, is underlined at 4: 458 and 5: 55: ‘By thinking itself into a world of understanding practical reason does not at all overstep its boundaries … causa noumenon … is, though a possible thinkable concept nevertheless an empty one.’ In the Grundlegung (4: 462) the restrictions are forcefully and vividly repeated in an echo of the anti-Platonic ‘dove’ imagery of B8–9:
… even though I have an Idea of it [the intelligible world], yet I have not the least cognizance of it … Here … is the highest limit of all moral enquiry and it is already of great importance to determine it so that reason may not, on one hand, search about in the world of sense for the supreme motive and a comprehensible but empirical interest, and may not, on the other hand, impotently flap its wings without moving from the spot in the space, which is empty, of transcendent concepts called the intelligible world, and so lose itself among phantoms.
Kant there underlines the defects both of merely empiricist and of pseudo-rationalist, transcendent accounts of morality. It is not surprising, then, to find him repeating the same restrictions in the second Critique 5: 57:
… as for whatever other properties, belonging to the theoretical way of representing such supersensible beings, may be brought forward … these are without exception to be counted not as knowledge but only as a warrant (for practical purposes a necessity) to admit and presuppose them … so … pure theoretical reason is not given the least encouragement to fly into the transcendent.
(b) It may be suggested that the restrictions imposed on a theoretical use of reason in this way nevertheless give way, according to Kant, to a practical knowledge of freedom and moral agency in which we ascribe moral features to things in themselves in an intelligible world. In the passage just quoted Kant talks only of a ‘warrant’ in presupposing an application to the super-sensible world, but this raises a question about the character of such a warrant. I claim that Kant is not committed to endorsing moral judgements about an intelligible world, and outline the structure of the argument from the resolution of the Third Antinomy to the applied moral philosophy.
What Kant claims to have established in his resolution of the Third Antinomy is, as is by now well known, severely limited. At B585–6 he denies even aiming to establish the ‘reality’ or the ‘possibility’ of (transcendental) freedom, and these restrictions are repeated in the later moral philosophy, for example, Grundlegung 4: 455: ‘Freedom is only an Idea whose objective reality is doubtful.’ What the arguments are designed to establish is only the ‘compatibility of freedom and nature’ (B586) in a position Kant importantly calls ‘defensive’ (4: 459). They show, as he says (5: 93–4), that ‘no insight can be had into the possibility of freedom as an efficient cause’ but only a ‘sufficient assurance that there can be no proof of its impossibility’. The point is repeated in the Grundlegung 4: 459:
Freedom … is a mere Idea the objective reality of which can in no way be presented in accordance with the laws of nature … Now where determination by laws of nature ceases there all explanation ceases as well, and nothing is left but defense, that is, to repel the objections of those who pretend to have seen deeper into the essence of things and therefore boldly declare that freedom is impossible.
That limited, defensive, strategy demonstrates the boundary between what is for us knowable and what is not (4: 458, 462). It entitles us to conceive of the ascription of both natural causality and transcendental freedom to one and the same being, but it does not explain any further the link with our actual moral responses and judgements. What is additionally needed is an argument from morality to give substance to, or ground for, our ‘interest’ (4: 459) in realizing the Idea of (transcendental) freedom in morality and the law (5: 93–4, 43). The point is clarified in the second Critique (5: 105) where the ‘principle of morality’ provides that transition from a merely ‘thinkable possibility’ (‘can’) to an actuality (‘is’):
Therefore that unconditioned causality and the capacity for it, freedom, and with it a being (I myself) that belongs to the sensible and to the intelligible world … is not merely thought indeterminately but is determined with respect to the law of its causality and cognized assertorically … which for theoretical purposes would be transcendent [but] is for practical purposes immanent.
Such a development in the realization of transcendental freedom still does not entitle us to claim to know the intelligible world theoretically. In particular, I believe, it does not entitle us to apply our moral judgements to such a world rather than to the world of sense viewed under the assumption of transcendental freedom. The point can be made by comparing passages from William James and Kant. James, in what he calls his ‘crass supernaturalism’, gave an account of a spiritual, religious world in the following terms: ‘The world interpreted religiously is not the material world over again with an altered expression … it must have a natural constitution different from that which a materialist world would have’ (James Reference James1902: 518).
The suggestion is that such a spiritual world has to be accepted as distinct from the material world of the senses in a literal ‘two world’ view. Kant, by contrast, seems to express a different, and more integrated, relation between the world of the senses and our moral judgements at B835–6 and in the second Critique (5: 43). There he reiterates the idea that the ground for an ‘intelligible’ world rests on the argument from morality, and goes on to characterize the upshot in our ascription of morality. He says:
The Idea of a moral world has, therefore, objective reality, not as referring to an object of an intelligible intuition (for we are quite unable to think any such object), but as referring to the sensible world viewed as being an object of pure reason in its practical employment. (B836)
This law is to furnish the sensible world, as a sensible nature (in what concerns rational beings), with the form of a world of understanding, that is of a supersensible nature, though without infringing on the mechanism of the former. (5: 43)
In Kant's account the upshot of the defence of freedom is only to integrate our moral, intelligible and our sensory, material judgements; it is not to divide or separate the two realms. Those two ways (standpoints) of talking (conceiving) are required to resolve the Antinomy, and minimally to defend a transcendental, super-sensible freedom, but it is recognition of the moral law which then allows us to talk immanently, as we do, of a human sensible nature subject to the demands of morality. Kant's arguments in his defence of transcendental freedom may not work, and his appeal to the fact of moral law may itself be flawed. But the consequence, in terms of his commitment to an intelligible world, is not very different from that in the theoretical context in which unschematized categories call on an intermediary schematism in order to realize them in a usable form in immanent experience. The difference is that we have an ‘interest’ in appealing to a super-sensible freedom, as a defence against determinism in the Antinomy, while we have no theoretical interest in appealing to a super-sensible world of unschematized categories.
4. Response to Adrian Moore
Adrian Moore and I have discussed his topic extensively and it is clear that there is very little agreement, or even convergence, in our views. We agree (how could we not?) that the mathematical antinomy rests on a systematic ambiguity between appearances and things in themselves. But even here Adrian Moore appeals to a reading of a separate passage (B768–9) which seems to me to be at best dubious. He uses it to make a contrast (which he admits is not described in this way by Kant) between the ‘ill-conceived’ mathematical and the ‘well-conceived’ dynamical antinomies (especially the third). The former's questions rest on confusion over the ambiguity noted above but the latter's do not. But, apart from the fact that in the passage Kant does not mention the Third Antinomy at all, it is surely wrong to think that that antinomy does not also rest on the same ambiguity. The unresolved dispute between determinism and freedom may not make a conscious contrast between appearances and things in themselves (any more than the mathematical antinomies), but Kant's resolution does rest essentially on that distinction. It allocates determinism to the realm of appearances and (transcendental) freedom to the realm of things in themselves.
That provides one, minor, example of the disagreements between our views, but I note some other more significant disagreements.
4.1
There is some misunderstanding of my position in the dispute over the references to ‘pure reason’ and ‘pure (a priori) reasoning’. Adrian Moore objects to the claim ascribed to me: ‘(1) It is a mistake to think that we can settle issues about the character of the physical universe by pure reason (671, 674).’ One misunderstanding arises from an omission in the cited passages, since there I talk of relying on pure (a priori) reason/reasoning alone. Once that omission is corrected the counterexamples of the proofs in the Analogies and mathematics (p. 236, above) have no relevance. In neither case are these examples of pure reason/reasoning alone; in the former a reference to possible experience, and in the latter to intuition, is required.
A second misunderstanding arises from Adrian Moore's account of ‘settling’ the issues in the Antinomies, whether through pure reason or not. The claim ascribed to me does not distinguish between ‘settling’ the dispute by choosing one of the opposed options and ‘resolving’ the dispute by showing why it cannot be in that sense ‘settled’. The issue can be ‘resolved’, according to Kant, precisely by showing why it cannot be ‘settled’. This confusion accounts, I think, for the complaint on p. 237, namely that I believe that the underlying mistake in the dispute ‘cannot be eradicated’ or ‘cannot be eradicated by pure reason’. Nothing in my account makes or entails such a claim. With the proviso that for Kant dialectical illusions are unavoidable, my view is that according to Kant the mistakes can at least be demonstrated and eradicated in that way. In what follows, and to avoid the noted confusion, I shall use ‘settle’ for the adoption of one or other side in the given dispute and ‘resolve’ for Kant's ‘higher order’ monitoring of that dispute.
There is an important corollary of Adrian Moore's failure to make clear that distinction between ‘settling’ and ‘resolving’. He disregards, or plays down, the difference between engaging in the dispute as a participant and Kant's acting as a higher order ‘umpire’ through what is called a ‘higher judicial reason’ (B767). This encourages him to represent Kant as accepting the participating arguments (in some way), and especially to endorse the later ‘Observations’ on those arguments. In an important example of this Adrian Moore cites a comment on the thesis of the Third Antinomy (p. 238), and ascribes the view to Kant. My general view is not that these later comments contain nothing that Kant might endorse but only that we should not assume that Kant endorses them. In this particular case there is a decisive reason not to take the comment to express Kant's own view, since it makes the claim that (transcendental) freedom ‘is thereby proved’ even though in his resolution of the Antinomy Kant expressly denies that claim (B585–6).
4.2
A further objection to my view (on Kant's behalf) that the physical universe cannot be investigated by ‘a priori reasoning alone’ disregards the reference to Kant's general diagnosis of dialectical illusion in such passages as B85–8, B186–7, B335–6, in the Amphiboly, the section on Phenomena and Noumena, and throughout the Dialectic. It is a diagnosis which emphasizes especially the misuse of logic, and its a priori reasoning, in a (pseudo-rational) metaphysics which aims to establish truths about a reality such as the physical universe. This is how, in line with that diagnostic background, Kant represents the role of such pseudo-rational inferences in the Antinomies as well as the Paralogisms (B397). Its effect on the mathematical antinomies is partly, as Adrian Moore accepts, to produce a confusion, an amphiboly, between talking of the physical universe as an appearance or as a thing in itself.
The misuse of logic, however, lies not only in overlooking that ambiguity but also in using formal logic not properly as a mere ‘canon’, but improperly as an ‘organon’, that is, purporting to establish a truth about some object. The consequence, in both arguments, is the attempt to infer material truths about the physical universe from purely formal, a priori reasoning and from concepts and reason alone (B87–8). It would be reasonable to conclude that when Kant pledges himself for the ‘correctness’ of the arguments (Prol 4: 339) he accepts that each participant successfully shows contradictions in the opposition. This is a correct use of logic as a ‘canon’, that is, as a test for consistency (or inconsistency). To go beyond that to draw the more positive conclusion in each case would be to use logic illicitly as an organon, a mistake compounded by the systematic ambiguity in the conceptions of the physical universe either as an appearance or as a thing in itself. The consequence is that the disputants end up effectively, and confusedly, talking as if the physical universe were a thing in itself, though this is simply outside our cognitive range. The issue between the disputants is then unsettlable or undecidable, but the resulting problem for a healthy reason is resolvable just by making that point.
4.3
That summary of Kant's position discloses a fundamental divergence between Adrian Moore's view and mine. One way of putting that difference would be to say that for Adrian Moore it is the very concept of the (unqualified) ‘physical universe as a whole’ which sends the debate into the undecidable realm of things in themselves, whereas for me it is the confusion of supposing that truths about the physical universe (as a whole or not) can be obtained through pure reason and a priori logic alone. It seems to me that Adrian Moore's alternative is open to two serious difficulties. The first arises if we ask why that unqualified concept of the physical universe as a whole should lead us into the undecidable realm of things in themselves. Adrian Moore's answer is that it rests on Kant's belief that the physical universe as a whole is simply not given in any possible experience. Kant does make such a claim about what he calls the ‘absolute’ conception of the physical world as a whole (B510–11), but this is in a context in which he distinguishes that notion from an ‘empirical and comparative’ conception of the physical universe as a whole.
One difficulty for Adrian Moore then arises from Kant's general criterion for what is real in the realm of appearances, namely that it makes a necessary reference to empirical (scientific) laws. The claim is that what is real ‘stands in connection with a perception in accordance with the laws of empirical advance’ (B521, B273–4). With that criterion Adrian Moore's conclusion that the (unqualified) physical universe as a whole (whether ‘absolute’ or only ‘comparative’) is a thing in itself would have to establish not only that it is not given in any possible experience, but also that we cannot connect it with a (given) perception in accordance with empirical laws in a respectable scientific cosmology. He equates ‘experience’ with ‘empirical cognition’ in n. 11 with the aim of avoiding this difficulty, but I cannot see that it succeeds. The note still does nothing to accommodate Kant's wish to allow as ‘reality’ whatever can be inferred from perception in accordance with empirical laws even in cosmology.
4.4
A second difficulty arises from that apparent deficiency. For it might be said that it is not feasible to meet Kant's further criterion, since we have no reason to deny the possibility of a physical, scientific cosmology with its empirical laws. Such a discipline might entitle us to draw legitimate conclusions about the physical universe as a whole, but these conclusions would refer to the physical universe as an appearance and not as a thing in itself. The current development of such a scientific, empirical cosmology lends added weight to this possibility. I cannot see that either Kant's resolution of the antinomy or Adrian Moore's version of it offers grounds to reject scientists’ current claims about the temporal origin of the physical universe. Indeed in my version of Kant's position he has no intention of offering such grounds.
What might still be said, however, is that even if we accept such a scientific development we have no good reason to think that Kant either did or would have accepted it as possible. But this seems to me wrong, as I indicate in my discussion in TRK. It seemed to me not only that Kant's resolution of the antinomy allowed the possibility of a scientific cosmology but that his own strategy throughout the Dialectic pointed in just that direction. Alongside Kant's rejection of specific pseudo-rational disciplines, such as rational psychology, he typically allows legitimate disciplines of an empirical kind. Why should cosmology be any different? Kant prepares for a similar development in his discussion of the antinomy when he distinguishes a transcendent sense for ‘world as a whole’ with its reference to an ‘absolute’ beginning or totality or indivisibility from its empirical, comparative and phenomenal sense (B510–11). Later, in section 9 (B543), he specifically envisages such an empirical use in the title, ‘The Empirical Employment of the Regulative Principle of Reason in Respect of All Cosmological Ideas’. Such a legitimate empirical employment for a principle of reason is outlined in the following pages, but it is only regulative and not constitutive. It does no more than encourage us to extend our investigation of the beginning of the physical universe, of the physical universe as a whole, as far as possible without giving any ground to think that we can complete such a task. Any conclusions from such a discipline would have the typically revisable character of any scientific theory. I do not see why current scientific cosmology should not be content with that.
4.5
These last comments indicate directly the alternative outcomes in Adrian Moore's and my accounts of the Antinomy, but there is a final, more general, correlate of his position which I still cannot accept despite the fact that he has weakened his view over the course of our discussions. In earlier versions he ascribed to Kant a straightforward knowledge ‘of’ things in themselves, but now he speaks only of knowledge ‘concerning’ things in themselves which is admitted to be ‘extremely tenuous’. Despite this proviso he nevertheless still says: ‘This relates to something else that I think Kant thinks we know: that there must ultimately be something unconditioned corresponding to whatever is conditioned (Bxx)’ (pp. 238–239). I do not think that such a claim is compatible with what Kant says, in the context of the Antinomy, of the unconditioned and our limited knowledge only of the conditioned in a possible experience at B510–11. Nor do I think that the references to Bxx, Bxxvi and 5: 196 are adequate to support the claim.
Because I deal with both Bxxvi and the moral parallel to 5: 196 in the response to Manfred Kuehn the crucial case is that from Bxx. In that passage, however, Kant explicitly and repeatedly denies knowledge of the unconditioned as a thing in itself. He says that although things in themselves are ‘real per se’ they ‘are not known by us’ and ‘we find that the unconditioned is not to be met with in things so far as we know them’. In the light of these strong qualifications to Kant's commitment to things in themselves, and particularly in line with the account of that commitment in the response to Manfred Kuehn, it seems to me that we need a more subtle account than a ‘tenuous’ appeal to the (transcendent) ‘way things really are’. Such an account, outlined in the response to Manfred Kuehn, would include the following: the necessary conceivability of things in themselves coupled with our principled ignorance of them; the wholly ‘defensive’ appeal to (transcendental) freedom in the resolution of the Third Antinomy; and the ‘interest’ of reason in making determinate the realization of freedom and the moral law within the sensible world of appearances.
5. Response to Ken Westphal
Ken Westphal has provided the clearest and most extensive account of my views in TRK and notes his agreement with almost all of them. He, of all the critics, seems to me most obviously to be counted as a ‘revolutionary’, or at least non-traditionalist, interpreter of the first Critique. That point is underlined further by the fact that even the reservations which he notes do not, I believe, constitute significant differences between us. Those reservations centre on his objection to a view I take about the relation between the first Critique and Kant's subsequent Metaphysical Foundations of Natural Science expressed in the paragraph quoted (pp. 248–249, above) from TRK, 486. The primary objection is that I accept some of Michael Friedman's views, which Ken Westphal thinks mistaken, so my response tries to deal with those crucial points in his account.
5.1
Ken Westphal's main anxiety (pp. 248–249) is that in the cited paragraph a role which both Michael Friedman and I assign to Metaphysical Foundations cannot be correct. However Friedman and I do not agree on the point which Ken Westphal regards as objectionable. Moreover my disagreement with Friedman (403 and 802 n. 11) which Ken Westphal also cites (n. 8) constitutes the very point that Ken Westphal himself makes against Friedman's view (p. 250). Michael Friedman's claim is that the fundamental principles of Newtonian science are, for Kant, transcendental conditions of the possibility of experience generally. My view is that the general transcendental conditions of experience outlined in the Analytic of Principles differ from the a priori, but not transcendental, principles which belong to Newtonian physics. As a consequence I take it that the roles of the Critique and Metaphysical Foundations are different, though certainly connected. When Ken Westphal insists (p. 250) that the Critique (but not Metaphysical Foundations) ‘must explain and justify our commonsense identification of, and causal judgements about, spatio-temporal objects and events prior to and independently of … natural science’, I entirely agree. The more general transcendental conditions of experience are the distinctive target of the Critique.
5.2
The suggestion is that my position in the cited paragraph is inconsistent, but the problem may arise more from carelessness of expression than from inconsistency. In the paragraph my intention was to note the possibility, implicit in the Analytic of Principles, of showing how the transcendental principles are realized in the particular context of physics. That those same principles are also realized less technically in the context of our common, non-scientific, experience is crucially important but consistent with that other more technical use. The two contexts represent just the difference between the transcendental principles which make experience generally possible, and the a priori, but not transcendental, principles which belong to Newtonian physics. That distinction respects Kant's contrast in the Analytic of Principles (B199–202) between the philosophical principles underlying science, mathematics and physics, and the mathematical and physical principles of those disciplines. Michael Friedman's view, in contrast to mine, depends on acceptance of a Quinean naturalism which I do not accept and which I believe Kant could not accept. These points were made in the symposium in the Aristotelian Society (Bird Reference Bird1998; Friedman Reference Friedman1998) where Friedman and I discussed the issue (see TRK, 820 n. 11).
5.3
Ken Westphal connects that issue about Metaphysical Foundations with a wider point about ‘empirical criteria’ for, and ‘impure’ a priori aspects of, the categories and their transcendental principles. He rightly notes that Kant talks at some length (B249–51) of the ‘empirical criteria’ for categories such as ‘cause’ and ‘substance’; and he helpfully refers to Konrad Cramer's work. Even in that extended passage from B249 to B251, however, Kant makes it clear that the discussion of empirical criteria is strictly supplementary to the pure a priori account of the categories and their principles. The point is made at the end of the passage at B252 where a distinction is drawn between the (empirical) content of ‘actual forces’ and their (a priori) form. The suggestion seems to be that, despite the inevitable connection between pure (a priori) and impure (empirical) aspects of the principles, there is a need to mark out a strictly a priori treatment of them in a formal Kantian metaphysics. This seems to be a motivation behind Kant's proviso about empirical criteria for ‘substance’ at B232.
5.4
These provisos undoubtedly raise questions about the status of pure a priori (transcendental) principles, and it may be that there remains some disagreement between us about this. In TRK I tried to distinguish a pure account of the categories from their partial and full realizations in common and in technical scientific experience. I also insisted on treating Kant's strictly a priori formal account of categories and principles as a theoretical, and somewhat artificial, abstraction from their actual deployment in common experience and science (see Kant's comments on Euclidean geometry in the Prolegomena 3: 374). Those distinctions seem to me especially important in contexts where Kant's formal apparatus has encouraged commentators to treat his abstract terms in the formal theory as if they referred to actual but mysterious, probably transcendent, objects. So a transcendental self may come to be regarded as a mysterious actual self, somehow lying behind the mere appearances of our everyday experience of ourselves and others; a ‘transcendental object’ may similarly come to be regarded as an equally mysterious thing detached from experience of tables, chairs, thoughts and pains. To interpret Kant's abstract formal theory in those ways is, I believe, to disregard Kant's warnings (B88) against confusing his formal doctrines with a bogus material interpretation (see the response to Adrian Moore). Those misunderstandings are typical of certain forms of ‘traditionalism’ and it is consequently particularly important to resist them not least because Kant's own philosophical therapy rests on a diagnosis of just such errors in what he derides as ‘windy metaphysics’. I would be surprised, in the light of his positive comments, if Ken Westphal disagreed with this.