At the outset of my career, I was summarily assigned the task of teaching the Critique of Pure Reason. Early on, I came across two books that helped me, and eventually my students, immensely: Peter Strawson's The Bounds of Sense (Strawson Reference Strawson1966) and Graham Bird's Kant's Theory of Knowledge (Bird Reference Bird1962). I took away from Strawson the idea that Kant's project had to do with determining the conditions of objective judgement and from Bird the idea that satisfaction of these conditions entailed a ‘realist’ interpretation of our experience of a world existing independently of that experience. When it came time to write my own book, Kant's Theory of Science (Brittan Reference Brittan1978), I titled it in such a way as to acknowledge my debt to Bird in particular.

Now beginners and old hands alike have his wonderfully helpful TRK to draw on. It is comprehensive, insightful, engaging. But the task at hand is more to criticize than it is to praise.

In the Introduction to the Critique (covered in Part I, sections 3 and 4, of TRK), Kant sets out the problem he intends to solve and the framework in terms of which that problem is posed. The statement of the problem is straightforward – how are synthetic a priori judgements possible? – as is the pair of distinctions which it presupposes – between analytic and synthetic judgements, on the one hand, and between a priori and a posteriori ways of coming to believe that they are true or false, on the other. This said, there has been endless controversy about the way in which the problem and its accompanying framework are to be understood.

There are two main difficulties with the framework and hence with Kant's statement of his problem. One, more traditional, has to do with distinguishing between analytic and a priori judgements in a sharp and significant way. All analytic judgements, which Kant variously characterizes in terms of containment and consistency, are a priori. But so too, the traditional difficulty runs, all a priori judgements are analytic. For the only way in which one could know independently or in advance of experience that a judgement is true depends on whether or not its predicate is somehow contained in its subject or its negation is self-contradictory. It matters little that ‘a priori’ and ‘analytic’ have different grammars, for both pick out all and only the same set of judgements. If they are thus co-extensional, it is not possible for a judgement to be a priori and yet not analytic.

Bird has a number of interesting things to say about this empiricist line of argument. Two of them are central. The first is that it begs a fundamental question in simply asserting that the only way in which one could know in advance or independently of experience that a judgement is true depends on whether or not the predicate is somehow contained in the subject or its negation is self-contradictory. Necessity and universality are, Kant says, the marks of the a priori. But the empiricists refuse to recognize any necessity except that which is de dicto; for them necessity is a property of sentences, not of the relations between things and their properties. But this is at least arguable. Moreover, as Bird points out, one of Kant's ‘Copernican’ aims is to show that the whole notion of ‘truths about the world’, independent of the ways in which we describe or otherwise structure it, is philosophically bankrupt.

All that follows is that it is possible to make a distinction between the analytic and the a priori, not that it is necessary to do so. Bird's second, more implicit, line of thought reinforces this conclusion. It is that the distinction is no more than proposed, and then rather hurriedly, in the Introduction to the Critique. Only later and in full detail is the argument set out that there are judgements which are both synthetic and which can be established a priori, the clearest examples of which are developed mid-way through the text in the Analogies of Experience.

The second, more recent, difficulty with Kant's framework was made famous by Quine. It is not that there is one distinction rather than two to be made between kinds of judgements, but that there is none. The analytic/synthetic, a priori/a posteriori lines are blurred at best. Quine's more important arguments for this claim may be set out briefly, although in a manner that is as much mine as it is Bird's.

First, there is no non-circular, non-question-begging way to distinguish analytic from synthetic sentences. If one says, as many do (although, importantly, not Kant himself), that analytic sentences are logical truths or are reducible to same upon substitution of synonyms, we are left with the problem of characterizing ‘synonyms’. If we characterize synonyms in terms of ‘sameness of meaning’, we are left with the problem of characterizing sameness of meaning, and so on.

Second, the analytic/synthetic distinction goes hand in hand with what Quine calls the ‘dogma of reductionism’, that synthetic sentences are confirmed or falsified by the sense experience sentences derivable from them. Analytic sentences are ‘vacuously confirmed, ipso facto, come what may’. But as any number of failed attempts to do so indicate, there is no non-circular, non-question-begging way to characterize the sets of sentences describing or reporting sense experiences or to relate such sets, deductively or inductively, to other sentences which they are held to confirm or falsify, a result apparently reinforced by the thesis of the indeterminacy of translation.

Third, the analytic/synthetic distinction suggests along the same lines that certain judgements are recalcitrant to experience and therefore never need to be revised in light of that experience. But there are no sentences, analytic or not, recalcitrant to experience; our judgements face its tribunal collectively, and (sometimes in the interests of simplicity) we should be prepared to abandon any of them, even so-called logical truths, in order to square that experience with our theories about it.

Bird rejects all three of these arguments. Quine's first argument trades, obviously enough, on the conventional assimilation of the a priori to the analytic, and does not enlist any additional considerations in its behalf. This assimilation depends in particular on an extensionalist point of view on which whatever differences there might be between the meanings of ‘analytic’ and ‘a priori’ can be safely ignored. Moreover, the charge that the characterization of the set of analytic sentences is eventually circular is inconclusive, for the same thing could be said about the characterization of any set of sentences or its corresponding concept.

Quine's second argument also trades on the conventional assimilation of the a priori to the analytic, for it is precisely Kant's point that there is a set of synthetic sentences, those that are a priori as well, that does not reduce to sentences describing or reporting sense experiences. Bird might have added, as he does in Kant's Theory of Knowledge, that no synthetic a posteriori sentences, insofar as they are about objects, can be so reduced. Kant is not a phenomenalist, indeed not a ‘reductionist’ of any kind. Finally, the thesis of the indeterminacy of translation, however dubious on other grounds, does not undermine in any serious way the distinction between analytic and synthetic judgements that Kant makes.

Quine's third argument, about the revisability of all of our judgements, again has more to do with the distinction between the a priori and the a posteriori, but is, all things considered, the most persuasive way in which to criticize Kant. For it has seemed to many people that those claims about the world he sought to establish as certain have been overthrown by subsequent developments in mathematics and science, from which it follows by an easy induction that the quest for certainty that he sought is futile. Bird contents himself with saying that however true this might be, ‘it does not follow that there is no other way of distinguishing them’ as analytic or synthetic, and therefore that here as well Quine's argument is inconclusive. Well and good, but the issues here deserve more discussion.

Perhaps no summary is needed, but I think Bird's way with respect to both traditional and recent criticisms of the analytic/synthetic distinction has three main components: that these criticisms assimilate without adequate argument the a priori to the analytic, that it does not follow from the (presumed) fact that the judgements in question are co-extensive that they cannot meaningfully be distinguished, and that, at least for the purposes of argument, we should grant Kant the distinction and see what he makes of it, notably in the case of those fundamental judgements that he claims are both synthetic and a priori.

The second of these replies preoccupies Bird, for he goes on at some length to talk about how the various distinctions can be ‘separately understood’ and to give ‘definitions’ of the concepts involved in such a way as to do so. For reasons that will emerge shortly, I am not very happy with these ‘definitions’. More to his point is that Kant's framework, however vulnerable to philosophical attack, can be used to pose a very deep and ostensibly legitimate problem: on the assumption that there are synthetic a priori judgements, on what does their ‘possibility’ depend? The task is twofold: to make good on the assumption and to answer the question.

In the Introduction to the Critique, Kant seems to think that it is enough to instance some propositions of pure mathematics and pure physics to make good on the assumption and so to set the stage for his question about their ‘possibility’. Bird disagrees. In his view, ‘It is commonly, and rightly, said that the famous argument to show that “7 + 5 = 12” is synthetic is fallacious’, and then, typically, he adds that whatever fallacies it involves do not entail that the judgement in question is analytic. Rather, he counsels us to wait until the Analytic of Principles before we pass judgement, so to speak, on the possibility of a synthetic a priori, sets out an analogy with the rules of chess, and (understandably enough, from my perspective) falls back on Kripke's celebrated discussion of related issues.

The chess analogy, a kind of Wittgensteinian survival in Bird's thought, is easily set out. Just as the rules of chess structure a particular activity and in the process define roles for the various pieces on the board, so too certain a priori principles structure our experience and in the process define roles for our various cognitive faculties. Both are constitutive of their corresponding activities, both are in this sense ‘a priori’. In particular, since the a priori principles structure our experience, they cannot be derived from or reduced to it. It is as though empiricism wanted to play tennis with the net down.

Bird understands the several limitations of the chess analogy. Its rules are entirely arbitrary, those in terms of which we structure experience are not. There would be no rules without rule-makers, we human beings, but rules that govern the ‘game of life’, so to speak, are conditionally necessary on the identity of self and the coherence of moral responsibility. Assuming these latter, there is no way in which we could structure our theoretical experience or our practical activities than spatially, temporally, causally, substantially and dutifully. At the same time, the analogy underlines the objectivity that the rules make possible, i.e. a particular move conforms to the rules or not independent of what transpires in a player's head, and the fact that they must be generally shared before the game can begin.

My approach to Bird's discussion of Kripke will be by way of my reservations concerning his characterization of both the framework and the central question it allows us to pose. To this point, Bird has argued that synthetic a priori judgements are ‘possible’ in a logical sense, since the analytic/synthetic and a priori/a posteriori distinctions do not necessarily collapse, and in an analogical sense as well, since we have examples, as in chess, of (normative) ‘judgements’ that are constitutive of and hence irreducible to experience. But the way in which the case has been made for the first two levels of ‘possibility’ prejudices the kind of case that should be made for the third and deepest level. Fortunately, Bird is already familiar with much of what follows and will understand that I could think of no better way to engage him.

We can begin with Bird's characterization of the analytic/synthetic distinction. It is in terms of types of truths. I know that Kant himself talks this way at times, although, so far as I can tell, never in the Introduction. At A151/B190, he says that the truth of analytic judgements ‘must always be cognized sufficiently in accordance with the principle of contradiction’. But this saying does not cohere very well with his view that ‘truth … is the agreement of cognition with its object’ (A58/B82) and that ‘an analytic [judgement] takes the understanding no further, and since it is occupied only with that which is already thought in the concept, it leaves it undecided whether the concept even has any relation to objects … what the concept might pertain to is indifferent’ (A258–9/B314).

This conclusion is reinforced by another passage in the Introduction. At B15, Kant writes:

On my reading, this means that analytic principles may be used in mathematics only on condition that their subject terms refer (‘exhibition in intuition’ being a sufficient condition of such reference). What we have to say, I think, is that for Kant an analytic proposition is necessarily true, that is, if its terms refer then the truth of the proposition can be determined via the law of contradiction.

Indeed, this idea, that the ‘validity’ of an analytic judgement does not entail its (knowable) truth seems to gird much of Kant's criticism of the ontological argument. ‘God is omnipotent’, he says at A595/B623; ‘that is a necessary judgement’. Since ‘necessity’ here is understood in terms of the inconsistency of its negation, it is analytic as well. But it is not (knowably) either true or false; no judgement whose subject term purports to refer to a super-sensible object is. There is (knowable) truth or falsity only when the subject term of a judgement refers to an object that can be given to sensibility, i.e. that can be located in space and time.

This point has a number of corollaries. The most important is that an analysis of concepts alone can never serve to determine reference, and hence, by the argument already given, recourse to intuition must also be made.Footnote 1 This is the point at which Kant breaks decisively with Leibniz. It is also the pivot on which the Critique turns. This is what Bird misses in his discussion of Kripke. He has him right on the distinctions between synthetic and analytic, a priori and a posteriori (and for that matter, between necessary and contingent). He understands the way in which these distinctions, and the particular use Kripke makes of them, is tied to the notion of rigid designation. But he does not see (or at least mention) the fact that Kripke invokes the notion of rigid designation in large part because he rejects descriptive (or we might also say, ‘analytical’) theories of reference.

This is the deep point of intersection between Kripke and Kant. When the latter defines the Critical project by the question raised in the famous letter to Herz of 1772 (‘on what basis [does] that in us which is called “representation” refer to an object?’, a question we might rephrase in a more linguistic mode as: how is it possible that singular terms, particularly those that have no empirical elements, manage to refer to objects that exist independently of our perception and conception of them?), it is on the assumption that a merely descriptive account of reference, such as that essayed in the Dissertation with respect to merely intelligible objects, is inadequate and must be replaced. The first half of the Critique is devoted to setting out an adequate theory of objective reference, the second half to demonstrating that many hallowed metaphysical judgements contain terms that do not satisfy it, and are hence neither true nor false so far as we humans are concerned.

Like most commentators, Bird takes what Kant says about mathematics in the Introduction as merely illustrative. In my view, not so. Kant's strategy is to begin with what might appear to be the most difficult case for his thesis that all objective reference requires intuition, the case of mathematics. For pure mathematics seems to turn entirely on the analysis and manipulation of concepts. If he can show that this is not true, that even the terms of mathematics are not genuinely singular and hence refer uniquely only if intuition is invoked, then he will have gone some way in the direction of persuading us of the general thesis that there is no determination of, and consequent reference to, any object whatsoever without recourse to intuition. This is to say that the case of mathematics supports the general thesis. The concepts of ‘seven’ and ‘five’ do not determine what ‘that single number may be which combines both’ (B15; my emphasis), that is to say, do not determine the unique reference of ‘twelve’.

I also think that Bird's way with ‘a priori’ is somewhat problematic, although most of the problems spring from Kant's multiply ambiguous use of the expression. Since my point remains generally the same, that Kant's central concern is with the conditions of objective reference, and in this sense with the ‘possibility’ of various kinds of judgement, I can be much briefer.

On Bird's definition, ‘a judgement is a priori (true) if and only if it does not require presented experience’. This definition leaves it open what an a priori concept or intuition might be,Footnote 2 and it is not made clear what additional premises are needed to go from ‘does not require presented experience’ to ‘holds universally and necessarily’, the twin marks of the a priori. But more than that, it suggests that ‘a priori’ characterizes a particular relation between claim and evidence, and is therefore epistemic and adjectival, rather than the role that certain concepts, intuitions, and judgements play in structuring experience, in which case it is primarily ontological (for want of a better word) and adverbial.

In fact, Kant uses ‘a priori’ in a number of different ways. Intuitions are a priori if they do not result from ‘affection’ by an object, concepts are a priori if they contain no empirical elements, judgements are a priori if they cannot be ‘derived’ from sense experience. All three are also a priori and therefore ‘possible’ in a much deeper sense: if certain concepts did not apply, certain intuitions were not available, certain judgements were not true, then experience itself (which I have been construing in terms of objective reference) would not be possible, in which case the concept of the self as subject of that experience would not have application.

Of course, Bird goes on in his discussion of the Analytic of Principles to spell much of this out, very clearly and insightfully. But it is perhaps worth making more explicit what I find troublesome in the ‘does not require presented experience’ definition of a priori.

First, the definition (taken at face value) assumes that ‘experience’ is unproblematic, when it is Kant's point (as Bird well knows) that even ‘presented experience’ is very much in need of philosophical analysis. The analysis, in turn, reveals that ‘experience’, in order to be experience of objects (and thus to satisfy the requirements of the continuing identity of the self) requires that certain synthetic judgements (or in the case of space and time certain pure concepts and intuitions) are a priori true of it.

Second, these concepts, intuitions, and judgements which must hold a priori of our experience of objects are precisely those which make objective reference possible. In his very able discussion of Kripke, Bird says that the notion of rigid designation ‘play[s] no part in Kant’. But it does. For Kripke, rigid designation is a feature of certain kinds of terms, those which are genuinely singular (pick out the same individual in all possible worlds). With the possible exception of ‘thing in general’, Kant does not think, as against Leibniz and Kripke, that there are any genuine singular terms; intuitions are always required. But his concern is the same as Kripke's, to identify the conditions with respect to which designation is possible. It turns out that these conditions hold not in or of all possible worlds, but of all really possible worlds, worlds in which we human beings are capable of making reference to objects. Put in other words, there are for Kant no terms which apply to all and only the same objects in all possible worlds, but there are certain conditions, expressed in judgements that are both synthetic and apply a priori, which are true of all and only the worlds we are capable of experiencing.

Bird may find it ironic, given that I have (over)emphasized points of disagreement between us, that my reading of the Introduction, indeed of the entire Critique, is rooted in reflection on his work and takes off from his realist perspective. No Bird, no Brittan.