Theism as a theory

A first crucial issue for the Bayesian apologetic strategy is whether theism really can be interpreted as an explanatory theory or hypothesis, which might be confirmed by empirical evidence. Theologians who opted for the via negativa have argued that God is so totally different from everything he created that we cannot describe his nature by employing any positive predicates. Other theists hold that if we specify the nature of God by using terms such as ‘person’, ‘free’, ‘powerful’, ‘knowledgeable’, or ‘good’, we are using these terms analogously, or as metaphors, since in their literal senses they apply to humans only.Footnote ⁴ I agree with Swinburne that if this were true, theism would not be a theory or hypothesis which can be (dis-)confirmed by empirical evidence. As he stresses correctly: ‘[i]f theology uses too many words in analogical senses, or loosens up the rules for their use too much, it will convey virtually nothing by what it says’ (Reference Swinburne CT , 75).Footnote ⁵ Accordingly, Swinburne argued in The Coherence of Theism that most predicates in terms of which his version of theism is defined are used literally.

In chapter 7 of Philipse (Reference Philipse2012) I argued in extenso, however, that by claiming that God is ‘without a body’, Swinburne annulled the conditions for applying literally and meaningfully the other terms by which he defined God. The stipulation that God is bodiless would imply a contradiction with these other defining terms, unless the latter are interpreted analogically. For example, God is said to be a person. In the ordinary and non-analogical sense of the term ‘person’, persons are human beings, and humans are corporal beings. That our bodily nature is implied by the term ‘person’ follows from many of its uses. When I say, for example, that I cannot attend your birthday party in person, what I mean is that I cannot be physically present. Consequently, the term ‘person’ is used analogously if one claims that a person is bodiless. Since Swinburne did play the ‘analogical card’ with regard to the term ‘person’ as applied to God in another respect, we should conclude that there is more analogy in the claim that God is a person than he admits.

In his positive argument to the effect that the bodiless-claim does not annul the literal use of ‘person’ in his definition of ‘God’, Swinburne assumed that when we say that a person ‘has a body’, this expresses a relation between a person and a body, which meets five conditions. Because, allegedly, one might imagine coherently that these conditions do not obtain, he concludes that the term ‘person’ is used literally when one defines God as a person without a body (Reference Swinburne CT , 112–114). As I argued, however, Swinburne misinterprets the expression ‘has a body’ in these contexts. When I say that ‘Marcus has a muscular body’, I do not express a relation between Marcus and his body, of which I can imagine coherently that it might not obtain. Rather, this type of phrase indicates that the property I am attributing to Marcus is a bodily property, and not a legal or social one, for example. Similarly, when I say that Mary has a powerful mind, I am not implying that Mary has a relation to an entity, her mind, of which I can imagine coherently that it does not obtain under certain conditions. What I express by a phrase of this type is that the property I am attributing to Mary is a mental characteristic. Hence, in his positive argument to the effect that the term ‘person’ is used literally in defining God as a person without a body, Swinburne misinterprets statements of the form ‘X has an F body’. If one denies that God has a body, the claim that he is a person has to be interpreted analogically through and through.Footnote ⁶

Indeed, there is much more analogy in Swinburne's characterization of God than one can allow if theism is to be a meaningful theory that can be tested empirically. As I spelled out (Philipse (Reference Philipse2012), 97–102), we typically use behavioural criteria for applying psychological terms to human beings, and these criteria specify in part the meaning of these terms. For example, it would make no sense to say that some father loves his son if we went on denying that the father ever displayed any of the wide range of behaviours and facial expressions that function as criteria for attributing love. Wittgenstein concluded correctly from an extensive analysis of our uses of psychological predicates that ‘only of a living human being and what resembles (behaves like) a living human being can one say: it has sensations; it sees; is blind; hears; is deaf; is conscious or unconscious’ (Wittgenstein (Reference Wittgenstein and Anscombe1953), I, §281). The same holds for words such as ‘knowing’ or morally ‘good’, in terms of which Swinburne defines his notion of God. It follows that all these words are used analogously to a very large extent if one adds the claim that God is a being ‘without a body’. One thereby annuls most of the criteria for applying these terms literally. As Anthony Kenny concluded from a similar analysis, ‘when we talk in the language of the divine metaphor, we do not really know what the metaphors are about’ (Kenny (Reference Kenny2004), 45).

In The Coherence of Theism, Swinburne raised two objections against this analysis of psychological terms, according to which behavioural criteria are essential to their literal meaning. First, he concluded from the fact that ‘quite clearly a person may on occasion give no public expression to her mental states’ that this might ‘always be the case for some person’ (Reference Swinburne CT , 115). As I argued, however (Philipse (Reference Philipse2012), 103), in drawing this conclusion Swinburne commits a fallacy of hasty generalization. With regard to rules of language, it is not true that what is sometimes possible is always possible. For example, it makes sense to say that a person is altruistic although he did not show this on a particular occasion. But saying that he is altruistic and never shows it in his behaviour, is either false or does not make any sense. We have to conclude again that by claiming that God is bodiless, Swinburne cancels most criteria for applying psychological terminology to God literally.

Second, Swinburne contends that ‘a spirit can give expression to her wants or fears, hopes, preferences, and so on, even though she lacks a body’ (Reference Swinburne CT , 115). For example, ‘[i]f she fears that a certain man will make a wrong choice, she may make marks on sand conveying a message to him’ (Reference Swinburne CT , 115). Clearly, however, the phrase ‘make marks on sand’ must be interpreted analogously in this sentence if the subject is said to be a spirit without a body, since making marks on sand in the literal sense is a bodily activity. I am at a loss to understand what ‘making marks on sand’ would mean if one adds that the subject lacks a body. The same conclusion holds for nearly all activities that theists attribute to the alleged referent of the term ‘God’.

On the basis of these and other arguments I concluded in the seventh chapter of Philipse (Reference Philipse2012) that theism cannot be an empirically testable theory.Footnote ⁷ Since most of the terms used in characterizing the god of theism are used analogously to a very large extent, they ‘will convey virtually nothing’ about the theoretical content theism is supposed to have (Swinburne (Reference Swinburne CT ), 75). The illusion that there is such a theoretical content arises if one notices insufficiently that the claim according to which God is bodiless annuls the literal meaning of the other adjectives in terms of which the nature of God is defined. This is my ‘Objection One’ against Swinburne's apologetic strategy.

The predictive power of theism

If the arguments indicated in the previous section are conclusive, theism cannot be a meaningful theory or existential hypothesis. Consequently, end-node options d, e, and f of the decision tree for religious apologetics are ruled out for theists. Perhaps not everyone will be persuaded, however.

This is why I use a strategy of subsidiary arguments, which functions as follows. Let us suppose, counterfactually, that most terms in the definition of theism are used literally, as Swinburne holds, so that Objection One is false. Consequently, theism would have a reasonably well-defined theoretical content. In order to be an empirically testable theory (option f), however, a second condition has to be fulfilled. It should be the case that the phenomena or empirical facts adduced as evidence for God's existence are more likely to occur if theism is true than if theism is false. In other words, theism as a theory should have at least some ‘predictive power’. I shall now argue that even if it were a meaningful theory, theism would not have any predictive power in this sense. This ‘Objection Two’ against Swinburne's apologetic strategy is subsidiary because it is of relevance only if Objection One is unconvincing. Let me first explain intuitively why it is implausible to assign any predictive power to theism, even on the counterfactual assumption that it has a well-defined propositional content. Subsequently, I shall examine critically Swinburne's method of doing so.

We are now assuming counterfactually that we know what it means to say that God as defined by Swinburne exists. If so, the hypothesis of theism, that God exists in fact, will have some predictive power with regard to a specific item of empirical evidence e ₁ only if we can plausibly attribute to God an intention to cause or bring about e ₁. Since God is defined as both omniscient and omnipotent, he will always achieve what he intends to do. Hence, the probability that God has or had an (all things considered) intention to bring about e ₁ will determine the likelihood that e ₁ exists given the hypothesis of theism. This likelihood may be expressed by the formula P(e ₁|h&k), in which ‘h’ symbolizes the hypothesis of theism combined with the probability that God intended or intends to cause e ₁, and ‘k’ represents background knowledge.Footnote ⁸

Our question is how to determine this latter probability. Since we are playing the game of natural theology, reference to an alleged divine revelation such as the Bible has little relevance. Such an appeal risks a petitio principii if it presupposes that God caused the revelation. Another option would be to say that this likelihood or probability P equals 1 for any e ₁ by definition (the range of possible values of P being defined as 0 ⩽ P ⩽ 1), because in his theistic definition of ‘God’ Swinburne includes the clause ‘and the creator of all things’. Indeed, he claims that ‘theism is a theory of very wide scope purporting to explain the universe and all its characteristics’ (Reference Swinburne EG , 108). However, this option would reduce apologetic strategy (f) to a merely verbal move. Furthermore, Swinburne stresses that ‘God is supposed to be perfectly free’ to create or not.Footnote ⁹ This implies that the creation-clause in his definition of ‘God’ should be interpreted as expressing a conclusion of the Bayesian strategy, and not as its premise. Given God's perfect freedom, then, how can we determine the probability that God intended to create at all, rather than not to create? And if we attribute to God an intention to create, how can we assess the probability that he intended to produce the present world rather than another possible world, which does not exist?

This problem of predictive power is pressing for theists because their god (God) is defined as absolutely perfect in all respects, whereas clearly the existing universe is not perfect. Consequently, attributing to God the intention to create our universe implies that we also attribute to him the intention of diminishing the average perfection of all existing entities taken together. Isn't it implausible to assume that God as defined would have such an embarrassing intention? According to Aristotle, who developed a sophisticated version of theism, God's eternal activity consists in thinking about his own thoughts, and God's perfection excludes his ever creating anything.Footnote ¹⁰ Should we not conclude that attributing to God an intention to create is nothing but an anthropomorphic projection? We humans have to create, because we are vulnerable finite beings with many needs and longings. Since God as defined by theism is perfect and unique, however, he cannot have any needs. Hence, there seems to be no good reason to assume either that he would intend to create anything at all, or to attribute to him an intention to create particular things.

How does Swinburne solve this problem of theism's predictive power? He first argues for a version of moral (meta-ethical) realism or objectivism, according to which normative moral statements are true or false, and if true, are true necessarily.Footnote ¹¹ Second, because God is defined as omniscient, it follows that God knows all moral truths. From God's perfect freedom, Swinburne then derives the conclusion that God will always opt for what is the morally best action defined in terms of these moral truths (or, if there is no best action, for one of the set of good actions). Since according to Swinburne we humans also know many of these moral truths, we would have at least some epistemic access to God's intentions. In other words, there are objective values of P(e _1−n|h&k) for all instances of empirical evidence e _1−n for or against the hypothesis h of theism, and we humans are able to estimate quite a few of these values reasonably well. For example, Swinburne claims that he can evaluate objectively the probability that God would have intended to create a world in which humans would exist, and holds that this probability is about 0.5 (Reference Swinburne EG , 120, 123).

What should we think of this solution to the problem of theism's predictive power? Let me avoid discussing here the plausibility of Swinburne's meta-ethical supervenience-realism.Footnote ¹² Realism or objectivism is a minority position in meta-ethics, and it has been criticized on many different grounds by a large number of authors from J. L. Mackie onwards. My approach is a different one. Let us assume (counterfactually again) that Swinburne's meta-ethical realism is correct. According to this view, moral propositions ‘strongly supervene’ on factual ones; they can be true or false; quite some of them are true; and, if true, they are true necessarily in a metaphysical sense (Swinburne (Reference Swinburne2015) ). Does it follow from such a version of moral realism that on the basis of our moral knowledge we humans have access to at least some of the moral norms that would guide God (if he exists), as is the case according to Swinburne? I do not think so, for the following reasons.

I agree with Swinburne that fundamental moral principles can be formulated as specifying which moral property supervenes on actions sufficiently defined by a number of non-moral predicates. If so formulated, moral principles have the following propositional structure: ‘All actions having non-moral properties N_{1, 2, . . .} are morally good’ (or bad, right, wrong, obligatory, permissible, recommendable, etc.). My first and trivial point is that given the meaning of the term ‘good’ (etc.), a moral proposition of the specified supervenience-form is equivalent to a prescriptive moral statement of the form ‘One ought to perform actions with non-moral properties N_{1, 2, . . .}’ (or ought not, etc.), where ‘N_{1, 2, . . .}’ includes a description of the type of situations in which one ought to commit the specified action. In other words, each supervenience-statement of a moral metaphysical necessity is equivalent to a moral prescription.

Second, to make this equivalence explicit is important for the following reason. Whereas in the supervenience-formulations there need not be any reference to the set of subjects for whom the moral truth holds, there is such a reference in the prescriptive statements. The latter ones invite us to wonder which set of moral subjects the term ‘One’ (or ‘Thou’ etc.) refers to in each specific moral prescription. This can be explicated easily. For example, in the moral prescription according to which ‘One ought to care for one's children’, the set of subjects indicated by ‘One’ clearly is the set of parents. With regard to ‘One ought to honour one's mother and father’, the set consists of children, that is, of all humans. The prescription ‘Thou shalt not commit adultery’ addresses married human beings, while the moral imperative ‘Thou shalt not covet’ is directed to those who can long to possess things that they do not possess.

My third and central point is that, clearly, God as defined by theism (assuming the definition makes sense) does not, and cannot, belong to the set of moral subjects to which any of the moral prescriptions we know of applies. Theism is a version of monotheism. Accordingly, God is not only perfect and without needs, but also unique in his kind. He is not a member of a group or tribe or biological species, whereas all moral prescriptions we know of have the function of enhancing the cohesion and flourishing of human groups, and of fostering the well-being of our congeners (cf. Mawson (Reference Mawson2002) ). It follows that it is a fundamental mistake to attribute intentions to cause or create things to God on the basis of the moral rules we know of. If it makes sense at all to assume that God is a member of the set of subjects addressed by any moral prescription, we should admit that we humans don't have a clue about the contents of the moral rules that would apply to the monotheist god, and which God would apply to himself. It follows that Swinburne's attempt to provide theism with predictive power fails dramatically. Any particular attribution of a likelihood to P(e _i|h&k) as defined above will result from anthropomorphic projections on the god of theism. Objectively speaking, such ascriptions are completely arbitrary.

This diagnosis is confirmed by Swinburne's attributions of creative intentions to God. In The Existence of God, Swinburne barely discusses the general issue of why God would create anything at all, and his answer to this crucial question is both confused and unconvincing.Footnote ¹³ Let me discuss here his attribution of specific morally motivated intentions to God only. Some of these alleged divine objectives are instrumental in the sense that they are said to depend on other ones. For example, since ‘[h]umanly free agents need bodies’, God would intend to create a physical universe if he intended to create humanly free agents (Reference Swinburne EG , 130). To the question why God would want to create agents or spirits at all, Swinburne answers that God ‘must surely do so’, since a ‘solitary God would be a bad state of affairs’. The reason is that ‘God needs to share, to interact, to love’ (Reference Swinburne EG , 119). Hence, Swinburne attributes a probability of 1 to the hypothesis that God intends to create ‘rational beings other than a single divine person’, that is, rational beings other than him, and a probability of 0.5 to the hypothesis that God intended to create morally free human beings (Reference Swinburne EG , 123).

Why should we think of ascribing these intentions to God? Surely, human beings need to share, to interact, and to love. We are social animals that would not have survived as a species if we lacked an urge to do these things. Without any doubt, the moral rule that ‘One ought to share, to interact, to love’ applies to us humans. But why would this rule be relevant to a monotheist god, who is not a social animal, and who can exist eternally on his own, being perfect in all respects? Obviously, to assume that God belongs to the set of moral subjects to which the norm applies is an anthropomorphic projection.

The same holds for Swinburne's attribution to God of the intention to create human beings. Because we are not morally perfect, we humans have a type of freedom that God lacks, the freedom to choose between good and bad actions. Why would God deem it morally good to create persons having this hazardous type of freedom? According to Swinburne, the goodness of this type of freedom is obvious, because ‘[w]e think it a good gift to our children that they choose their own path in life for good or ill’ (Reference Swinburne EG , 119). But of course, from the fact that something is a good gift to our children, it does not follow that it is good to create children at all, even for human beings. Furthermore, if it is morally good for human beings to have children, it does not follow that it is good for God to create the human species. Applying these moral rules to God clearly is an anthropomorphic projection.

We cannot but conclude that if we want to abstain from committing the blasphemous mistake of attributing human and all-too-human intentions to God, we have to admit that we cannot legitimately ascribe any specific intentions to him. Hence, if theism were to make sense, it would lack predictive power.

Extending the chain of subsidiary arguments

One of Swinburne's great merits is that he elaborated his apologetic strategy in minute detail, anticipating many objections. Since I focus here on structural problems concerning this strategy, I shall skip details, and merely apply my scheme of subsidiary arguments a few more times.

Let us assume counterfactually, then, not only that theism is a meaningful theory, but also that it has predictive power in the sense defined (option f). Allegedly, we humans are able to estimate the probability that specific facts obtain given theism, because we have access somehow to moral truths that hold for God. Stipulating that both Objection One and Objection Two are false, I move on to Objection Three.

Swinburne claims that theism resembles a scientific theory because it is empirically testable, just like the hypothesis that there are quarks, or that neutrinos exist (Swinburne (Reference Swinburne1983), 386). Scientists and philosophers of science agree, however, that empirical testability in this sense requires that we can derive factual statements from the relevant hypothesis (in combination with other premises) of which we do not yet know whether they are true or false. What really counts, Lakatos once wrote, is whether a theory or hypothesis implies ‘dramatic, unexpected, stunning predictions’ and ‘leads to the discovery of hitherto unknown novel facts’ (Lakatos (Reference Lakatos, Curd and Cover1998), 25). If theism were a theory with predictive power, then, it should be possible in principle that it can be disconfirmed by factual discoveries, and that the total evidence turns out to be negative rather than positive. Real predictive power implies the risk of refutation. As we shall see, however, Swinburne refuses to run this risk for theism.

In the past, Christians have derived many factual statements from theism combined with moral insights. For example, assuming on moral grounds that God revealed the text of the Bible to us, countless Christian experts have attempted to calculate the date at which God created the universe on the basis of biblical genealogies. Numerous chronologies were published during the sixteenth and seventeenth centuries, of which Bishop Ussher's is best known. According to a recent exegesis, ‘what Ussher actually said was that Creation [of Heaven and Earth] took place during the evening of Saturday 22 October 4004 bc’ (Jackson (Reference Jackson2006), 24). Unfortunately, however, this theistic prediction has been refuted by many scientific results. For example, radiometric methods of age dating revealed that both meteorite material and the Earth are about 4.54 billion years old. Furthermore, Big Bang theory combined with measurements of the expansion rate of the universe show that our universe started to expand some 13.8 billion years ago.

Since during the history of Christianity a great many factual statements have been derived from theism that turned out to be false later on, we should diagnose this doctrine as a degenerating research programme in Lakatos's sense, if at least we want to interpret theism as a theory with predictive power (option f). In other words, the counterfactual assumption that my objections One and Two are false gives rise to Objection Three, according to which we should reject theism on empirical grounds.

Swinburne attempts to eliminate this third problem by restricting the predictive power of theism to facts of which we already know that they obtain. Refusing to run the risk of disconfirmations, he avers that the hypothesis of theism ‘does not yield predictions such that we can know only tomorrow, and not today, whether they succeed’ (Reference Swinburne EG , 70). However, by adopting this strategy of immunization against empirical refutation, Swinburne in fact gives up his claim that theism is an empirically testable theory, just like the hypothesis that neutrinos exist. Furthermore, he contradicts his allegation that ‘theism is a theory of very wide scope purporting to explain the universe and all its characteristics’ (Reference Swinburne EG , 108). This is Objection Four.

According to my fifth objection, Swinburne's strategy of immunization fails to be successful. As we saw, he estimated the probability that God would create humanly free agents at about ½, and since such creatures need bodies, the probability that God would create a physical world in order to enable their existence would be at least 0.5. Since he adduced some other reasons why God would create a physical universe, such as its beauty and God's intention to create animals, P(physical universe|h&k) would exceed ½. However, if we can attribute intentions to God on the basis of our moral intuitions, many recent cosmological discoveries convince me that the probability that God would create our universe equals zero.

Perhaps we can accommodate the discovery that life is extremely rare in the universe, both spatially and temporally, in contradistinction to what Christians thought when they embraced Aristotelian geocentrism. However, since cosmologists are now convinced that our universe will continue to expand forever, a so-called heat death of the universe may be inevitable. In such a state of thermodynamic equilibrium at a very low temperature, no life will be possible, and in the ‘dark era’ beauty is inconceivable. Since this state will last an infinitely long time, the fraction of time at which the universe enables life to exist would approach zero. Surely, a good God would never create such a universe.

Let me finish by mentioning three other structural problems for Swinburne's apologetic policy, which I explained at length in Philipse (Reference Philipse2012). First, Swinburne staged a second strategy of immunization, in order to exclude a priori any explanatory competition between theism and empirical science. He stipulated that the predictive power of theism should be limited to phenomena that are in principle either ‘too odd’ or ‘too big’ to be explained by secular science or scholarship (Reference Swinburne EG , 74). According to my Objection Six, this second strategy of immunization not only contradicts the claim that theism really is a testable theory. It also fails to be successful.

For example, in Swinburne's argument to the effect that Christ's bodily resurrection occurred and confirms theism, the primary phenomenon or evidence e to be explained is not this resurrection, which would indeed be ‘too odd’ to be explained by secular science. It consists rather of the alleged testimonies of the resurrection, which have been passed down to us in the New Testament. As I argued at length (Philipse (Reference Philipse2012), 166–182), a secular explanation of these testimonies is vastly superior to Swinburne's theistic hypothesis that the miracle really occurred and was caused by God. Furthermore, I showed that Swinburne's argument to the effect that the existence of the universe is ‘too big’ for science to explain even if the universe would stretch back in time infinitely, contains a fatal fallacy (ibid., 184–185). In fact, none of the empirical phenomena that Swinburne adduces as evidence for theism are in principle too odd or too big to be explained scientifically, if at least they really exist (ibid., 183–188, and passim).

My last two (sets of) structural problems for Swinburne's apologetic strategy are related to his use of Bayes's theorem, according to which the posterior probability of a hypothesis given evidence e, that is P(h|e&k), can be calculated by multiplying the prior probability of the hypothesis P(h|k) with the likelihood of the evidence given the hypothesis and background knowledge, P(e|h&k), divided by the likelihood of the evidence given background knowledge P(e|k):

What Swinburne intended to show in The Existence of God is that by applying Bayes's theorem repeatedly to various pieces of evidence e _i for or against theism, we can conclude that on the total evidence pro and contra, the existence of God is more probable than not, that is, that P(h|e _1–11&k) > ½. As I stressed (ibid., 311), applying this ‘cumulative case strategy’ is one of Swinburne's great merits. Since empirical arguments pro or contra a hypothesis may reinforce or weaken each other, we should study the total evidence and its interrelations if we want to reach a conclusion about the ultimate probability that the hypothesis is true.

One set of problems (Objection Seven) is concerned with the initial prior probability of theism P(h|k). If the value of this prior is arbitrary, that is, up to our subjective estimates, Swinburne's cumulative case for theism will not convince any unbeliever. They will estimate the prior of theism to be low indeed, and judge that the evidence Swinburne adduces is not sufficient to raise the posterior to above ½. Hence, Swinburne endorses an objectivist interpretation of Bayesianism, and holds that because theism is a very simple theory, its prior is not too low. I argued in extenso, however, that if we can determine the complexity or simplicity of theism at all, we must conclude that theism is much more complex than the doctrine that the Big Bang occurred as an ultimate brute fact, for example (ibid., 238–255, cf. 201–207). Furthermore, I pointed out (ibid., 212–220) that according to most philosophers of science, simplicity merely is a pragmatic criterion for theory choice rather than a truth-conducive property of theories, as Swinburne has to assume.

The last set of problems I shall indicate here (Objection Eight) is concerned with the denominator P(e|k). A piece of empirical evidence e confirms a hypothesis if and only if the relevance criterion is met: P(e|h&k) > P(e|k). In Swinburne's terminology, the occurrence of e provides a correct confirming or ‘C-inductive’ argument for h iff P(e|h&k) > P(e|k), that is, iff P(e|h&k) divided by P(e|k) is greater than 1, and it confirms h to the extent that this fraction exceeds 1. However, while formula P(e|k) of the denominator appears to be a simple one, it conceals great complexity. First, the following equivalence holds: P(e|k) = P(e|h&k).P(h|k) + P(e|~h&k).P(~h|k). Second, in this latter formula ~h functions as a so-called ‘catch-all hypothesis’, because, in principle, we may formulate infinitely many non-theistic theories that are competing with theism, since they aim at explaining the same evidence. On various grounds, I argued that the value of the denominator P(e|k) cannot be calculated, so that the probability of theism given the total evidence, P(h|e _1–11&k), cannot be computed either (ibid., 205–207, 220, 245–255, 311–317).

Swinburne's greatest and indeed monumental merit in philosophy is to have elaborated apologetic strategy (f) in favour of theism better than any philosopher of religion before him. Whether his Bayesian campaign for theism is conclusive is another matter, as will be clear from the fatal objections I formulated. I wish him success in refuting at least some of them.