Kant's thoughts about God are not easy to reconstruct. Not only are they spread across a wide variety of texts but those texts themselves were produced over a period of nearly twenty-five years. Given that over this period there were changes in Kant's terminology, in the focus of his interest, and in his views, it is not surprising that one can construct very different accounts of what he believed. At one extreme, he can be presented as accepting (or at least as being committed to accepting) a kind of atheism. For he appears to say that the claims about God made by traditional theism are senseless, a kind of transcendental illusion, and hence that ‘God exists’ must be false, or at least not true. At the other extreme, Kant can be seen as a theist. For in relation to our theoretical concerns, he claims that God as the ground of all being is something which every thinker must presuppose; and in our practical concerns, Kant famously offers a moral proof of God's existence in which God is portrayed as the presupposition of the necessary truths of morality, and hence surely as necessarily existing Himself.

In the research monograph Kant on God, Peter Byrne aims to steer a careful route between these two extremes. While acknowledging that there are plain inconsistencies in the relevant Kantian texts (they have a ‘tensive, paradoxical character’ (2)), he tries to construct from those texts a coherent account; and although he is concerned with exposition rather than assessment, it is an account which he seems to think has some plausibility. His conclusion is that Kant is closer to deism than to any form of conventional theism, and he thereby aligns himself with traditional interpretations of Kant, stretching from Heine in the early nineteenth century through to Guyer in our own day, which take Kant's position to be largely destructive of traditional theism. Correlatively, he distances himself from some more recent interpretations, e.g. by Wood, who claims that Kant was not fundamentally hostile to traditional rational theology, and by Hare and Palmquist, who take Kant to be a theist and indeed a Christian theist.

Part of Byrne's case for this account of Kant lies in his selection of the texts on which to focus. Thus, he attaches little weight to the Lectures on Rational Theology, on the grounds that they were constructed from notes by Kant's students and not seen to press by Kant himself. At the same time, he regards the Opus Postumum as a significant source for Kant's ideas, and very substantially in harmony with the earlier Critical corpus. Since the Lectures suggest a Kant fairly sympathetic to natural theology, and the Opus a Kant who is much more hostile, Byrne argues that the relevant texts support a significantly sceptical account of God.

The first three chapters of the book outline Kant's Critical Philosophy, and his discussion of theoretical arguments for the existence of God. A chapter on Kant's analysis of God talk is then followed by three chapters exploring the moral basis for Kant's acceptance of God. A final chapter is concerned with current discussions of whether Kant was, after all, a standard Christian theist, and argues that Kant is much nearer to eighteenth-century deism than to conventional Christianity.

The main distinctive feature of Byrne's account and the one that underpins his interpretation of Kant as radical revisionist, is the emphasis he places on Kant's account of the meaning of claims about God. He argues that Kant's reservations about our relationship to God are not merely epistemological (denying reason in order to make room for faith, etc.), but are also semantic. He first sketches this position in the following terms:

Byrne later offers a more detailed version of this claim, saying that Kant distinguishes between what he (Byrne) calls a ‘bare’ and a ‘full’ sense of the term ‘God’ (64). In the bare sense, ‘God’ means ‘the unknowable source or ground of order’, and it can be literally true in that sense that God exists. The full sense of ‘God’ is one which includes the range of properties which rational theology ascribes to God, and it is no more than ‘a useful fiction’ that God in the full sense has these further properties. In effect, traditional theology runs together the bare and full senses. So, Kant is presented as accepting the old semantics for ‘God’ as ‘the source of cosmic order, whatever that may be’, but as offering a new semantics for ‘God is an omnipotent, supremely just, loving, etc. creator’.

(We might note in passing that talk here of two senses for ‘God’ is misplaced. In both literal and metaphorical claims about God, ‘God’ must have the same sense. It is only because there is a unitary sense that assertions about him can be divided into the literal and metaphorical. (There isn't a metaphorical sense of ‘ship’ in which it is true that ships plough through water.) But I will overlook that point in what follows, and continue to speak of two senses.)

This interpretation certainly has significant textual support. It has the merit, for example, of making sense of an otherwise puzzling passage in the first Critique at A696/B724. Kant says that it certain that there is something ‘distinct from the world, which contains the ground of the order of the world and its connection in accordance with universal laws’; and then he immediately follows this by saying that claims that this something is a substance, is necessary, is the ens realissimum, etc. are all illegitimate. On Byrne's account, the first of these claims presumably uses ‘God’ in the bare sense, and the others are illegitimate if taken literally but legitimate if taken metaphorically.

The distinction between the two senses of ‘God’ is then put to work by Kant in three areas of our intellectual concerns. We have to refer to God in relation to our empirical investigation of the natural world, then more specifically in relation to our studies of the living world, and thirdly in relation to our moral thinking. Given Byrne's two-meanings account of Kant, we would thus expect Kant to say that in relation to these three areas, we are entitled to draw the literal conclusion that we can know there exists a ground or source for the natural order, for the teleological order and for the moral order, but that if we wish to ascribe to this ground any of the more specific predicates of traditional theism (omnipotence, omniscience perfect goodness, etc.), such claims will be legitimate only if taken metaphorically.

In fact, in a number of places, this isn't quite what Kant says; nor when commenting on Kant, does Byrne always say quite what his two-meanings account commits him to. For example, in relation to the ground of intelligible orderliness in the natural world, Byrne says that for Kant, ‘The postulation of God is no more than one way of articulating [the] assumption of a humanly intelligible universe’ (80, italics added). This interpretation certainly gives us a sense in which it is literally true that God exists (i.e. when ‘God exists’ means ‘the universe is humanly intelligible’), but what it abandons is Kant's insistence that there really (literally) is something ‘distinct from the world’, albeit a something which we can characterize further only metaphorically.

In relation to the second and third domains (teleology in the living world, and the harmony of virtue and happiness) Byrne's two-meanings account would commit Kant to saying that in each domain we can know that in a literal sense there is a ground for teleological interpretations of the living world and of the moral world, but that when we go beyond this claim and say, e.g., that the ground is a designer (hence a person with intentions, etc.), or that the ground is loving and perfectly just (and hence able to secure the harmony between virtue and happiness), we have moved to the realm of metaphor. With some forcing of Kant's text, he can be seen as doing this in the third Critique when he says that physical teleology affords evidence sufficient for our theoretical reflective judgment to enable us to admit the existence of an intelligible world cause.

Byrne judges both of these lines of thought, about teleology in the living world and about morality, to be unpersuasive, and that Kant must therefore rely mainly on the thought that our theoretical enquiries presuppose God.

It easy to agree with Byrne's condemnation of the teleological and moral arguments for the literal existence of a ground or source, but it is hard to agree with him that the argument for God as a presupposition of empirical enquiry fares any better. First, a minor point: Kant repeatedly says that we are forced (presumably on pain of irrationality) to postulate God, whereas Byrne's Kant wavers on this point, sometimes saying that the postulate of God is ‘no more than one way’ of thinking (see passage quoted above), sometimes accepting that we are forced to think this way. But leaving this point aside, we must ask what this postulate, whether unavoidable or optional, amounts to. On Byrne's two-meanings account of Kant, it must be a two-pronged postulate. The first prong forces us to accept as true the claim that God (in the sense of a ground for the orderly intelligibility of the world), exists; the second prong tells us that any further predicates legitimately ascribed to God can be used only metaphorically.

However, there is a cluster of problems here for Byrne's interpretation, at least if it is offering a coherent account of Kant. First, what could be meant by saying that we must presuppose that there is a ground or source to the intelligibility of the world? Given that we cannot say literally anything further about this ground, the assumption that it exists could not have any explanatory or heuristic function at all, in furthering empirical investigation. Of course in one sense, it is trivially true that there must be a ‘ground’ for this intelligibility. For the source of the concept of a ground lies in the relational judgement feature of being hypothetical. So it looks as if Kant is committed to saying that p is a ground for q if and only if there is some true conditional of the form ‘If p, then q’; and surely, for any true q, there will be some such conditionals. But if we try to attach a more restrictive interpretation to talk of grounds, for example thinking of the ground as being literally the cause of order, or even as the explanation for order, then Byrne's Kant claims that we are trying to use concepts beyond the limits of possible experience, and thus illegitimately.

Secondly, even if we admit that in the basic sense, there is a ground for the intelligibility of the empirical world, why must we add that in the full and metaphorical sense, this ground is a rational designer? We can agree that I can think of the empirical world as if it were the product of a rational designer, just as I can also think of it as if it were the product of many independent designers, or indeed of purely natural processes. Since I am not thereby thinking that any of these as-if thoughts describe how things actually are, I am not thinking of them as supplying any explanation of why the world is as it is, nor of why it should be intelligible to human investigators. Certainly, someone who engages in these as-if thoughts need not be making a mistake, but nor is someone who declines altogether such forms of thought.

These problems arise from Byrne's assumption that Kant has a theory of meaning in anything like the modern sense. Kant certainly speaks of some things as being meaningless, or as lacking significance, or as being empty, which may suggest that he has some ideas about where the boundary of meaning lies, and some theory about why it lies where it does. But on an alternative reading which fits the texts at least as well as Byrne's, Kant's concern is primarily epistemological. On this reading, the categories set bounds to knowledge but not to meaning. The thesis that knowledge requires intuitions and concepts, and that any concept requires the use of the categories gives Kant something like the conclusion that any knowledge acquisition or retention must use the categories.

The problem with categories or concepts for which there are no corresponding intuitions is not that the concept is meaningless (or more precisely, that a supposed verbal expression has no meaning) – it is that such concepts are empty, where ‘empty’ means ‘cannot be used to make any true knowledge claims’. Certainly, Kant says that such concepts are ‘impossible’ (e.g. in the first Critique A319/B178). But most commonly when he speaks of something as possible, he means so-called real possibility not logical possibility, where, ‘really possible’ means ‘can be known to be true or false’. So such passages do not support a semantic against an epistemological interpretation of Kant.

On this epistemological account of the Kantian enterprise, we are free to use concepts in general or categories in particular to have thoughts in relation to which we can never achieve knowledge, or gain any kind of empirical evidence. I can thus wonder whether there are Xs, or believe that there are Xs, or hope that there are Xs and so on, with perfect logical propriety. I am thereby stepping outside the limits of possible knowledge, and this leads Kant to describe such entities as not ‘really’ possible. But this means only that the preconditions for knowledge of such things are not met. It is only if we mistakenly attribute to Kant some kind of quasi-positivist connection between knowledge and meaning that we could take him to mean that talk of such things is semantically deficient.

Apply this line of thought to God and we have the following: the God of traditional rationalist theology is not an object of possible experience, is therefore unknowable, and is therefore not a ‘real’ possibility. Someone who claims knowledge of such a being has certainly made a mistake – but it is an epistemological not semantic mistake. While it is true that I cannot know whether the categories apply to God, equally I cannot know that they do not. The contrast between these rival interpretations could be put as follows. According to Byrne, Kant says: detailed predications about God taken literally are senseless, so they must be taken metaphorically. According to the epistemological reading, Kant says: detailed predications about God taken literally are unknowable, but should nonetheless be taken literally. So taken, they can be the object of hope, etc. even though we cannot know whether the hope is ever fulfilled. Kant postulates a duality of cognitive status, not a duality of sense.

In the comments above, I have concentrated (as reviewers tend to do) on areas where I think that Byrne's interpretation is one-sided or at least questionable. But it must also be mentioned how very rewarding the book is to read, and indeed to study. Byrne is completely at home in all the Kantian texts, mining them effortlessly for ideas, objections, inconsistencies, insights, etc. There is also much of interest in the book that I have not so far mentioned. Chapter 6, for example, has a detailed discussion of Kant on happiness, and its exact relationship to virtue; chapter 10 points to some very interesting parallels and contrasts between Kant and traditional eighteenth-century deists; and throughout the text, there are many richly suggestive and thought-provoking asides. At 173 pages of text, the book is not long. But its short length belies the importance of the contribution it makes to thought about Kant on religion. It is a distinguished addition to the Ashgate Studies in the History of Philosophical Theology.