Home » Library » Modern Library » Richard Gale Swinburne Argument

Richard Gale Swinburne Argument

Swinburne’s Argument from Religious Experience (1994)

The following article was originally published in Reason and the Christian Religion, ed. by Alan Padgett (Oxford: Oxford University Press, 1994).


I have long admired Richard Swinburne’s work, not only for the way it has raised the level of discussion in the philosophy of religion by the introduction of technical sophistication and rigour, but even more for its courageous honesty in espousing and defending to the hilt his deepest beliefs and convictions, regardless of whether they are currently in vogue. He is a true professor whose concern is not to look good but to seek the truth, and for this he deserves our deepest respect as a philosopher.

Of all his many important contributions to philosophy, the one for which he is most likely to achieve lasting fame is his empirical argument for the existence of God in The Existence of God, a book that will become a classic in my opinion.1 As a result of this work, a return visit from Hume’s Philo is needed, but he had better come loaded for bear, because the weapons that he used so effectively to stop poor Cleanthes in his tracks will be of no avail.

While Swinburne’s overall aim is to establish that the probability that God exists is greater than one-half, he does not want the probability to be too high, for he fears that this would necessitate belief in God on the part of whoever accepts the argument, thereby negating the accepter’s freedom to choose not to believe. ‘If God’s existence, justice, and intentions became items of evident common knowledge, then man’s freedom would in effect be vastly curtailed’ (p. 245). An ontological argument would do even greater violence to the traditional Christian view of God as wanting men to come to know, love, and obey him of their own free will. If someone were to come up with a really convincing version of the ontological argument, Swinburne might not be crushed if we followed the example of the Pythagoreans, who set adrift sans supplies the person who demonstrated the existence of irrational numbers. Swinburne, therefore, must tread a very narrow line indeed, possibly a vanishing one.

I believe that Swinburne has created a needless difficulty for himself. First, he has mislocated the point at which creaturely free will enters into the relationship between man and God. It does not concern the freedom with which we believe that God exists, for it is dubious that in general we can believe at will, be it for conceptual or only causal reasons. Our free endeavouring, rather, concerns the efforts we make to adopt and adhere to a religious way of life, part of which is doing that which will help to self-induce a belief that God exists. One could accept the ontological argument, as Russell once did, and yet have no inclination to follow a religious way of life.

Second, there is much more free play between belief and assessment of probabilities than Swinburne allows. He has claimed that ‘S believes that p if and only if he believes that p is more probable than any alternative’, in which the alternative normally is not-p.2 It certainly is possible for someone to believe a proposition while believing that it is improbable, even highly improbable. I need not go back to Kierkegaard or Tertullian for an example, since I can appeal to my own case as a young boy when I believed with all my heart that the Giants would win the pennant, even though I believed that their chances were extremely remote; I wouldn’t have bet my piggy bank on it. Furthermore, one can believe that one horse in a race with a thousand horses in it has a better chance of winning than any other horse but not believe that it will win, since its chances of doing so are slight. And this holds even when the context is such that the alternatives to this proposition are propositions asserting that some other horse in the race will win. If we are to bet on one of these horses, naturally we will bet on the one that has the best chance of winning, however, slight this, assuming that the pay-offs are the same; but this just shows that believing cannot be analysed in terms of readiness to bet. Swinburne’s analysis is hardly an analysis of our ordinary concept of belief, but at best of our concept of rational belief. Thus, Swinburne’s worry about giving too good an argument for the existence of God is unfounded. His real worry should be that he hasn’t given a good enough argument.

There is a beautiful architectonic to his overall argument, replete with a surprising ending. Up until chapter 13, in which he presents the argument from religious experience, he appears to be making a cumulative case for the existence of God being probable, based on an agglomeration of the premisses of many probabilistic arguments that severally appeal to a wide range of evidence: namely, that there is a world at all, that it displays widespread law-like uniformity and simplicity in its governing laws and theories, that there exist organisms and-in particular-conscious ones, that men have great opportunities for freely co-operating in gaining knowledge and shaping the universe, and that history displays certain meaningful patterns, including the apparent existence of miracles. This counters the divide-and-conquer strategy of the theologian who considers each of these arguments in isolation, and shows that alone it doesn’t amount to very much.

For each of the above arguments a Bayesian-based case is made out that the evidence appealed to in its premisses makes it more likely than otherwise that God exists, which he calls a good ‘C-inductive argument’. Using P for probability, h to stand for the proposition that the God of traditional theism exists, e for the empirical evidence contained in the premiss of the argument, and k for background knowledge, the claim is that, for each of these arguments, the empirical evidence, e, that it appeals to is such that P(h/e and k) > P(h/k). He also argues that the kinds and amounts of evil we find in the world do not form the basis for a good C-inductive argument for not-h, since they do not lower the probability of h.

After he has completed this task at the end of chapter 12, the reader wonders what will result when the empirical premisses of all these arguments are agglomerated. Let us call the resulting argument the ‘cumulative probabilistic argument’; the question is whether it renders h more likely to be true than not (called a good P-inductive argument). Using e’ to stand for the conjunction of all these premisses, the question is whether P(h/e’ and k) > 1/2. The reader is getting ready to respond by saying, ‘Who knows?’ Since each of the arguments is a C-inductive argument, it is only comparative; and because each fails to assign any numerical value to the probability of h, there is no procedure for determining what happens when we add them together.3 There can be several good C-inductive arguments for a proposition p which together do not bestow a probability of greater than 1/2 upon p. But the reader realizes that it would be premature to engage in this bottom-line assessment, since there is another argument to come, the one in chapter 13, from religious experience. But the expectation is that the addition to the list of yet another C-inductive argument for h will not dispel our uncertainty about what to say. Mush added to mush gives mush.

It is here that the argument takes an unexpected turn. Instead of making out the usual Bayesian case that the evidence consisting of numerous religious experiences renders h more probable than it would otherwise be, which would be yet another good C-inductive argument, it appeals to an a priori presumptive inference rule, the ‘Principle of Credulity’ (hereafter PC), which renders it prima facie probable that the apparent object of religious experience, God, exists, and which can be defeated only if our background knowledge makes it is very improbable that God exists. And it is here that the foregoing highly intuitive cumulative case is appealed to for the purpose of defeating this lone possible defeater, thereby rendering it probable without qualification that God exists. Thus, Swinburne’s overall P-inductive argument for the existence of God is as follows:

(1) For any person S and any object x, if S, on the basis of an apparent experience of x, takes it that x exists, then prima facie it is probable that S’s experience is veridical and thereby x exists. (PC)

(2) If any person S, on the basis of an apparent experience of God, takes it that God exists, then prima facie it is probable that God exists. (From (1) by universal instantiation)

(3) Numerous persons, on the basis of apparent experiences of God, take it that God exists. (Premiss)

(4) Prima facie it is probable that their experiences are veridical and thereby God exists. (From (2) and (3) by modus ponens)

(5) That it is prima facie probable that their experiences are veridical, and thereby the claim that God exists can be defeated only if it is significantly more probable than not that God does not exist. (Premiss)

(6) The cumulative argument shows that it is not significantly more probable than not that God does not exist. (Premiss)

(7) It is probable without qualification that their experiences are veridical, and thereby God exists. (From (3), (4), and (5))

It is the purpose of this chapter to critically evaluate this argument. It will turn out that the fatal step in the argument is the deduction of (2) from (1), the reason being that the PC applies only to perceptual-type experiences, and religious experiences, on conceptual grounds, fail to qualify as perceptual. Instead of faulting (2) with an improper universal instantiation of the PC, it could just as well be charged that the PC is not properly restricted in (1) to perceptual-type experiences. These objections come to the same thing.

For the purpose of evaluating this argument, we will restrict ourselves to non-sensory experiences that the subject takes to be ‘directly’ of God in the sense that, relative to our human epistemic circumstances-the manner in which we humans experience the world and process information-they involve a minimum amount of interpretation and inference, unlike the experience of some empirical state of affairs as God-caused, for example, a sunset as an expression of God’s love or an evil as a test set for us by God. The latter are the religious experience analogue of the highly ‘interpretative’ and theoretically laden perception of a track in a Wilson cloud chamber as a movement of an electron, while ‘directly’ perceiving God is a ‘recognitional’ perceiving-as similar to perceiving something as a chair.4

Before I press my objection to Swinburne’s argument, it must be further fleshed out, beginning with the PC in premiss 1. Swinburne describes this principle as follows: ‘I suggest that it is a principle of rationality that (in the absence of special considerations) if it seems (epistemically) to a subject that x is present, then probably x is present,’ in which an epistemic seeming is one that serves as a basis for the subject to believe that the apparent object of the seeming exists and is as it seems to be (p. 254). Does this PC apply to any epistemic seeming, so that if it non-perceptually seems to a person that some proposition is true, then probably it is, unless there is reason to think otherwise? Such an unrestricted form of the PC is quite dubious. However, it seems clear that Swinburne means to restrict it to perceptual seemings, for immediately upon giving the above formulation of the PC he adds this paraphrase of it: ‘what one seems to perceive is probably so. How things seem to be is good grounds for a belief about how things are.’ And when he later recaps his argument from religious experience he writes: ‘My conclusion about the considerable evidential force of religious experience depends on my Principle of Credulity that apparent perceptions ought to be taken at their face value in the absence of positive reason for challenge’ (p. 275; see also pp. 272 and 276).

Another indication that he intends to restrict the PC to perceptual seemings is that he applies it to religious experiences, which in turn are defined as perceptual. Initially he defines a ‘religious experience’ as one ‘which seems (epistemically) to the subject to be an experience of God…or some other supernatural thing’ (p. 246).5 But he then goes on to say that the person who veridically experiences God will ‘in the very general sense…perceive God’, and talks ‘of such awareness of God as a perception without implying that the awareness is necessarily mediated by any of the ordinary senses’ (p. 247). Herein Swinburne allows for the possibility of non-sensory perceptions. They alone will concern us, since religious experiences that have a sensory content are highly inferential.

Notice that Swinburne’s PC is a retail principle that applies piecemeal to individual perceptual-type experiences. This contrasts with William Alston’s wholesale approach, according to which any well-established social practice for forming objective beliefs on the basis of certain types of perceptual inputs, subject to certain overriders, which he calls a ‘doxastic practice’, is to be taken to be reliable unless its outputs are either internally inconsistent or inconsistent with those of better-established doxastic practices. Since a reliable doxastic practice has belief outputs that are generally true, it would follow that any one of its perceptually based beliefs is probably true, which is exactly what results from the piecemeal application to it of Swinburne’s PC. Alston claims superiority for his approach on the grounds that Swinburne’s principle applies to experience-belief pairs individually, in isolation, while in my approach a principle of justification that applies to individual beliefs is grounded in a defense of the rationality of socially established doxastic practices. This provides support for my position that is unavailable to Swinburne.’6 Swinburne’s retail form of justification is open to supplementation by Alston’s wholesale approach, unless Swinburne doesn’t trust people who tell him that they can get it for him wholesale.

Furthermore, both claim an a priori status for their respective innocent-until-proved-guilty justificatory principle, and see it as our only alternative to complete scepticism. According to Swinburne, any attempt to inductively justify the PC on the grounds that most past perceptually based beliefs are true would have to assume a version of the PC which applies to beliefs based on apparent memories, since only through memory could we have access to this inductive evidence. But, obviously, the latter version of the PC does not admit of any non-circular inductive justification, and thus must be accepted if complete scepticism is to be avoided. ‘And if it is justifiable to use it when other justifications fail in memory cases, what good argument can be given against using it in other kinds of case when other justifications fail?’ (p. 256). Alston’s justification is similar in that it is based on the impossibility of giving a non-circular justification for any doxastic practice, again presenting us with the alternative of accepting the PC, now in its wholesale form, or embracing scepticism.

The prima-facie probability that the PC bestows upon perceptually based beliefs is subject to various sorts of possible defeating conditions-conditions that lower the probability that the experience is veridical and that its apparent object therefore exists. A defeating condition is really a flunked test. It is essential that there be possible defeaters, for otherwise it would be meaningless to speak of prima-facie justification. Swinburne lists four generic defeating conditions: (i) the subject or conditions under which the apparent perception were made are of a sort that have proved to be unreliable; (ii) there is inductive evidence that it is not possible for the subject to have perceived what he or she claimed to perceive; (iii) background evidence shows that very probably the apparent object was not present; and (iv) there is evidence that the apparent object probably was not part of the cause of the perception.

Having clarified premiss 1’s PC, it can now be asked how Swinburne justifies its application to religious experiences in premiss 2. Immediately upon stating the PC, he instantiates it with religious experiences: ‘From this it would follow that, in the absence of special considerations, all religious experiences ought to be taken by their subjects as genuine, and hence as substantial grounds for belief in the existence of their apparent object-God’ (p. 254). His justification for so instantiating the PC is given at the end of the chapter.

Initial scepticism about perceptual claims-regarding them as guilty until proved innocent-will give you no knowledge at all. Initial credulity is the only attitude a rational man can take-there is no half-way house. However, claims which can subsequently be shown unreasonable can be weeded out. But the onus remains on the challenger. Unless we take perceptual claims seriously, whatever they are about, we shall find ourselves in an epistemological Queer Street. Religious perceptual claims deserve to be taken as seriously as perceptual claims of any other kind. (p 276)

Here we see the basis for a rearguard action argument from analogy with sense experience for the applicability of the PC to religious experiences. Having shown that we readily apply the PC to beliefs based upon apparent memories and sense experiences so as to avoid scepticism, he states: ‘And if it is all right to use it for other experiences, they need a good argument to show that it is not all right to use it for religious experiences’ (p. 254). Herein Swinburne is placing the onus on those who refuse to extend the PC from sense- and memory-based beliefs to beliefs based on religious experiences: show some relevant disanalogy that would justify such restrictiveness, or shut up. And he is committing himself to neutralizing any proffered disanalogy, which task occupies the remainder of the chapter. Herein the analogy between religious and sense experiences is not given up front but in the course of shooting down alleged disanalogies. It is my intention to accept his challenge, but first I must complete my fleshing out of his argument.

There is little to question in (3)-Numerous persons, on the basis of apparent experiences of God, take it that God exists-since the experiences of these persons are being described in a way that does not entail the existence of their apparent object, called an ‘internal description’ by Swinburne. He claims that ‘all arguments from religious experience must be phrased as arguments from experiences given internal descriptions’ so as not to beg the question (p. 245). To accept the reports of others to have had certain internally described experiences, we need not appeal to Swinburne’s Principle of Testimony-‘Other things being equal, we think that what others tell us that they perceived, probably happened’ (p. 271)-which employs an ‘external’, or existentially committed, description of their experiences. A weaker, internalized version of it- namely, that other things being equal, we think that what others tell us that they apparently perceived probably was apparently perceived by them-will suffice as justification for the acceptance. It need only be added that the other things are in fact equal, since there is no good evidence that all those who reported having apparent perceptions of God are untrustworthy. Wouldn’t you be willing to buy a used car from Saint Teresa?

Pace Swinburne, his strong Principle of Testimony is not needed to justify either the acceptance of (3) or the transferability of the evidential value that these religious experiences have for their subjects to others. To show that if S’s apparent perception of x gives S a prima-facie justification for believing that it is probable that x exists, then it bestows a prima-facie justification for anyone to believe likewise, appeal need be made only to my weakened version of the Principle of Testimony and the principle of the universalizability of being-evidence-for, which Swinburne spells out as holding that ‘if e is evidence for h, this is a relation which holds quite independently of who knows about e’ (p. 260). The subject of the experience is more certain of its occurrence than is the hearer who must take the subject’s word for it by appeal to my weakened Principle of Testimony, and thus there is, as Swinburne points out, some small decrease in the probability of the proposition relative to the hearer over what it has relative to the subject-‘if S reports that it seems (epistemically) to S that x is present, then that is reason for others also to believe that x is present, although not as good reason as it is for S if in fact he is having the experience which he reports’ (p. 274).

Since there is no mystery as to how (4)-prima facie it is probable that their experiences are veridical and thereby God exists-follows from (2) and (3) by modus ponens, we can skip to (5)-that it is prima-facie probable that their experiences are veridical, and thereby the claim that God exists can be defeated only if it is significantly more probable than not that God does not exist. Swinburne’s argument for (5) consists in attempting to show for each of the above four generic defeaters that the only way in which they could apply to religious experiences, given the actual facts about such experiences, is if it could be shown that it is significantly more probable than not that God does not exist.

The test set for religious experiences by defeater (i) requires that ‘Most religious experiences are had by men who normally make reliable claims, and have not recently taken drugs’ (p. 265). And this test is in fact passed by most religious experiences. It is unclear how their reliability in ‘normal’ eases, such as making claims based on ordinary sense experience, has any relevance to their competence or ability to have veridical non-sensory perceptions of God. The cases are too unlike, because we can specify what constitutes a normal perceiver and standard conditions for sense experience but not for religious experience, a point that Swinburne himself emphasizes (see pp. 262 and 269). God is a supernatural being who freely decides upon whom to bestow grace by making himself an object of their experience. What Swinburne really does is to argue that religious experiences cannot run afoul of (i) because the test it poses is conceptually inapplicable to them. A lot more will be said about this when an attempt is made to show cognitively relevant disanalogies between sense and religious experiences that destroy his rearguard analogical argument.

Defeater (ii) requires that normally claims based on religious experience are true. But the only way in which it could be shown that religious experiences fail to satisfy this requirement is ‘If there was a good proof of the non-existence of God…. But the point here is that the onus of proof is on the atheist; if he cannot make his case the claim of religious experience stands.’ The prima-facie probability it bestows on the claim to have perceived God is yet to be defeated.

Swinburne considers another way in which the veridicality of religious experiences might be impugned-by appeal to the conflicting claims that are made on the basis of religious experiences within different religious traditions. Swinburne’s strategy is to argue that the extent of the conflict has been exaggerated, and that for the most part these claims can be shown to be compatible, since God could choose to present himself under different guises to persons who are in different cultural circumstances. He even gets down to why it is reasonable that a Portuguese peasant would ‘see’ Mary as attired in the manner in which she is pictured on the walls of Portuguese churches rather than as she was in the old days in Palestine.

That seems to me to count not at all against the peasant’s claim. For if Mary has survived death, what reason is there to suppose that she has now to dress the way she did in Palestine? If she is to manifest herself in bodily form the obvious way for her to dress is the way in which she would be recognized by those to whom she appears. (p. 268)

This raises the intriguing question of where she does her shopping. I could have sworn that I saw her the other day in the clothing section of a K-Mart in Pittsburgh, and thus I suspect that there will shortly be reports of visions of Mary in this area. It will be seen that Swinburne fails to neutralize the challenge posed by religious diversity because he considers only apparent inconsistencies between the claims based on religious experiences within diverse religious traditions and not the seeming clash between their different tests for the veridicality of such experiences. Again, he fails to detect a damaging disanalogy that would justify not extending the PC from sense to religious experience.

The requirement set by defeater (iii) would be flunked by religious experiences only if it could be shown

that very probably God was not present to be perceived, and so the subject could not have perceived him. But if there is a God, he is everywhere. He is only not present if he does not exist. So to use this challenge, you have to prove that very, very probably there is no God, and, as stated above, the onus is on the atheist to do so. (p. 269)

The prima-facie probability that apparent perceptions of God are veridical is not defeated by the fact that most other perceivers fail to have similar experiences, the reason being the above stated one that

we do not know that all persons with certain equipment and concepts could be expected to have experience of God, if he was there… the fact that there are no obvious disconfirming observations (no observations of the absence of God) which could be made, has the consequence that the original perceptual claim is, on its own, somewhat less evidence of the existence of God. (p. 269)

But it does not lessen the claim so much that it becomes less than one-half, and thus its initial claim to being prima facie probably true is not defeated. Again Swinburne is really showing that some potential defeating test is conceptually not applicable to religious experiences, thereby raising doubt about their cognitivity.

He holds that if theism is true, religious experiences cannot run afoul of (iv), which requires in this case that they be caused by God. God is the ultimate cause of everything, and moreover is omnipresent, not in the sense of being physically present everywhere, which would make him a cosmic fat man, but in being able to bring about effects at any place, including a perception of himself by those upon whom he bestows this grace. ‘If there is a God, any experience which seems to be of God, will be genuine-will be of God’ (p. 270). Thus, religious experiences pass the test imposed by (iv) with flying colours, assuming theism.

This is much too quick. Many religious experiences are caused in ignoble or devious ways, as for example by the ingestion of LSD or unconscious sexual desires.7 What are we to say about such experiences? Some would see them as forming the evidential base for a good C-inductive argument for the non-existence of God. For, if God exists, he is their ultimate cause. But, given the causal theory of perception, it would follow that they are veridical perceptions of him; however, it would be inconsistent with his goodness to bring about veridical experiences of himself in such ignoble and devious ways. A more reasonable way of dealing with them is to say that, although it is true that if God exists he must be at least a necessary cause of them, it does not follow that they are veridical perceptions of God, the reason being that they are not caused by God in the right way. God, moreover, could have some morally exonerating excuse for permitting to be unveridical apparent perceptions of himself, such as that they were caused by the subjects freely ingesting LSD or not doing everything in their power to rid themselves of sexual lust. This allows for the possibility that a religious experience may flunk the being-caused-in-the-right-way-by-God requirement, even if God exists.8

Swinburne fails to see this possibility because his mistaken causal theory of perception-‘S perceives x (believing that he is so doing) if and only if an experience of its seeming (epistemically) to S that x is present was caused by x‘s being present’ (p. 247)-fails to include a caused-in-the-right-way requirement. The presence of a chair can cause my apparent perception of it without my perceiving it if, for example, it caused someone who was probing my brain with electrodes so as to deceive me to see it and as a result produce in me the experience of seeming to see a chair. The causal chain that goes from the presence of the chair to my subsequent seeming to perceive it is too kinky or devious. While Swinburne is mistaken in his claim that religious experiences, on the assumption of theism, cannot fail (iv)’s causal requirement, properly understood, no serious damage is done to his overall argument; for it is reasonable to say that there are many religious experiences that do not in fact flunk this requirement. There is no evidence that the great mystics were high on something or had minds filled with smut or the like. Of course, if God does not exist, then any apparent perception of him flunks the caused-in-the-right-way requirement. Therefore, if the prima-facie probability that this experience bestows on the existence of God is to be defeated, the onus again is on the atheist to show that it is significantly improbable that God exists.

Before considering whether premiss 6’s claim that the cumulative argument is strong enough to defeat this defeater, it must be asked with what right Swinburne makes the strong demand of the defeater that it is significantly more probable than not that God does not exist. Why wouldn’t it suffice if it is only probable that God does not exist? There are some, like Gary Guting, who would demand even less of the defeater, since they reject Swinburne’s strong version of the PC in favour of a weaker one according to which an apparent experience of x merely constitutes ‘significant but not sufficient evidence’ that x is present, hereby departing from Swinburne, who takes it to be alone sufficient to justify an assertion that x is present, since it bestows a probability of greater than one-half on x‘s being present.9 I agree with Guting’s objection, for the sake of argument I will accept Swinburne’s PC, and thus will confine myself to the objection that it is not required of the defeater that it is significantly improbable that God exists but only that it is improbable that he does.

To see how Swinburne justifies his strong demand, we must begin with his generic account of defeater iii that the apparent object x of S’s experience wasn’t present or existent. ‘It has to make it very improbable that x was present if it is to outweigh the force of S’s experience sufficiently for it to remain more probable than not that x was not present’ (p. 261).10 Because religious experiences have an apparent object that is qualitatively remote from the subject’s past experiences and do not admit of the possibility of disconfirming experiences, Swinburne lowers his demand on the defeater so that it need show only that it is significantly more probable than not, rather than very probable, that God does not exist. ‘But for these qualifications, I would have concluded that a religious experience apparently of God ought to be taken as veridical unless it can be shown on other grounds that very, very probably God does not exist’ (p. 270).

There certainly is nothing intuitively obvious about this. Consider the generic case in which Swinburne requires the defeater to show that it is ‘very improbable that x was present’. It would seem to be sufficient for the defeater to show that relative to certain background evidence e1 the probability that x was not present be equal to or greater than that of the probability that x was present relative to the apparent experience of x. And since the probability of x‘s being present relative to an apparent perception of it need be only more than one-half according to the PC, however slight that might be, the improbability that x was present relative to e1 need be only less than one-half, and thus not very improbable. Similar considerations apply, only more so, to the case of religious experiences.11

If this objection is sound, Swinburne’s premiss 5 will have to be weakened to something like

(5′) That it is prima facie probable that their experiences are veridical and thereby God exists can be defeated only if the probability that God does not exist relative to evidence e1 is equal to or greater than the probability that God exists relative to the apparent experiences of him.

This will require a corresponding change of 6 to

(6′) The cumulative argument shows that it is not the case that the probability that God does not exist relative to evidence e1 is equal to or greater than the probability that God exists relative to the apparent experiences of him.

The demands that are made upon the cumulative argument, therefore, will depend upon whether it must counter (5) or (5′). My objection to (5) is far from conclusive, and thus both cases must be considered. Any neutralizer of (5)’s defeater also functions in the same way for (5′), but not vice versa. Thus, (5) makes the weaker demand, and is more easily neutralized. There are two main grounds for doubting that the cumulative probabilistic argument is up to even the less demanding task.

First, its account of the a priori prior probability of h relative to the tautological background knowledge k rests on a radical abuse of the concept of simplicity. It is claimed that although the P(h/k) is very low, since for any contingent being, of which God is one, it is extremely unlikely that it exists, God is far more likely to exist than is any other contingent being. This is because the only relevant consideration in such a case of prior probability is simplicity, and God is far and away the simplest of all beings, the reason being that since God has all his perfections to an infinite extent it can’t be asked, as it can be for possession of any finite degree thereof, why it has that degree rather than some larger or smaller degree. ‘A finite limitation cries out for an explanation of why there is just that particular limit, in a way that limitlessness does not. There is a neatness about zero and infinity which particular finite numbers lack’ (p. 283).

Herein Swinburne is confounding the simplicity of a being or its manner of possessing a property with its not requiring an explanation. Certainly, from a mathematical point of view, infinity is more complex than finitude, since it is defined in terms of an operation upon finite or natural numbers. And from a conceptual point of view, God’s infinite possession of perfections is far more complex and troubling than is the finite possession thereof. It was not for no reason that Swinburne felt a need to write a book entitled The Coherence of Theism. God’s omni-properties are far more problematic and paradox-prone than are their finite counterparts. Conceptually speaking, God is the most complex and difficult to grasp of all possible beings.

Second, Swinburne fails to do justice to the challenge posed by natural evils to his cumulative probabilistic argument. He argues with extreme implausibility that the amount and distribution of natural evil does not even lower the probability of God’s existence over what it would otherwise be. Swinburne constructs a theodicy for these evils based on their being necessary for our having the requisite knowledge to make morally significant choices, a knowledge of which can be gained only by inductive reasoning from past experiences of actual instances of natural evils.

if men are to have the opportunity to bring about serious evils for themselves or others by actions or negligence, or to prevent their occurrence, and if all knowledge of the future is obtained by normal induction, that is by induction from patterns of similar events in the past-then there must be serious natural evils occurring to man or animals. (p. 211)

In response to the objection that God could simply tell us what are the consequences for good or evil if we make certain choices, he responds that by doing so he would make his existence and intentions so manifest that it would usurp our freedom, since we ‘would have little temptation to do wrong-it would be the mark of both prudence and reason to do what was virtuous. Yet a man only has a genuine choice of destiny if he has reason for pursuing either good or evil courses of action’ (p. 212).

There are two fatal flaws in this theodicy. First, it fails to realize that it is only a contingent fact, if it is a fact at all, that all human knowledge of empirical matters of fact is gained via inductive reasoning from past experience. Certainly, God’s omnipotence would enable him to instil in us innate knowledge of the good and evil consequences of different acts, and, moreover, to do so in a way that did not reveal his existence or intentions; it would then be up to us to freely decide what to do on the basis of this knowledge. That Swinburne fails to see this alternative is further evidence of his failure to properly locate the point at which creaturely free will is operative.12 The other flaw is that it fails to deal with specific instances of natural evil that do not seem to be necessary for our gaining the requisite knowledge for functioning as moral agents. In effect, his theodicy is little more than a defence which offers a possible exonerating excuse for God’s permitting a certain kind of evil without any attempt to show that this excusing condition actually applies to all known instances of this kind of evil.

For these reasons Swinburne has completely failed to show that the many actual cases of apparently gratuitous natural evils do not significantly lower the probability of God’s existence. According to the intuitions of many, a good P-inductive, no less than a good C-inductive, argument can be based on such evils. If the cumulative argument, when purged of its abuse of simplicity in determining the prior probability of h, is sufficiently strong, the conjunction of its premisses with a premiss reporting all known instances of apparently gratuitous evil could still result in a good C-inductive argument or, less plausibly, a good P-inductive argument.13 If we are to be fully honest, I think we must admit that we do not have clear-cut intuitions about what to say about this enlarged and purified cumulative argument, and, as a result, have no basis for saying whether it suffices to neutralize either the strong defeating condition of (5) or even the weaker one of (5′). Thus, we are no better off than we were at the end of chapter 12. We are still mired in the mush.

This completes my critical exposition of Swinburne’s overall argument. While many objections have been lodged against it, none is really decisive. An attempt will now be made to give such an objection based on its being illegitimate to extend the PC to religious experiences, as is done in premiss 2. In the course of making it, relevant disanalogies between sense and religious experience will be unearthed that will more than answer the challenge issued by Swinburne’s rearguard argument from analogy to either produce such disanalogies or permit the extension of the PC from sense to religious experience.

It has already been seen that Swinburne himself seems to confine the PC to perceptual experiences. But what is a perceptual experience? It will not do to say that it is an involuntary experience in which some object is presented to the subject, for certain paradigmatic subjective experiences, experiences that take cognate or internal accusatives and do not admit of the veridical-unveridical distinction, such as sensations of pain and after-images, along with dreams and certain kinds of feelings, share these features. Nor will it do to define it as ‘an awareness of something apart from oneself’, as Swinburne does (p. 247). I could be thinking of Jones, and thereby be aware of him without perceiving him. Fortunately, it is not necessary for present purposes to give a definition of ‘perceive’. It will suffice to say that the conditions specified by the PC are at least necessary conditions for being such an experience.14

According to the PC, an apparent experience of x renders it prima facie probable that x is present, subject to challenge by certain defeaters. Both a metaphysical and an epistemological requirement for being a perceptual experience can be extracted from this. The metaphysical condition requires that the apparent object of the experience admits of the possibility of ‘objective’ existence, meaning that it can exist unexperienced and is capable of being the common accusative of different perceptions. The epistemological condition requires that the experience admit of the veridical-unveridical distinction by appeal to background defeaters or tests.

It is crucial that my account of these conditions be generic, so as to apply to all species of perception, including non-sensory perceptions of God. Were it to be confined to the case of sense perception I could rightly be charged with illegitimately taking features of one species of perception as normative for all other species. But because sense perception is both a clear-cut and a paradigmatic case of perception, for each of these conditions I will first give an account of how it is satisfied by sense perception. Abstraction then will be made from sensory qualities or anything that entails having sensory qualities, thereby rendering the account generic. It will be seen that religious experiences satisfy neither of these necessary conditions and thus fail to qualify as perceptual, and, as a consequence are not subject to the PC.

The metaphysical requirement. The object of a veridical sense perception can exist unperceived and be the common object of perceptions by different perceivers at the same time and the same perceiver at different times.15 This is because the object, along with the perceivers, are housed in the dimensions of space and time within which they are hooked up to this object by different causal chains. The tests for the veridicality of a sense perception-having the right sort of causal hook-up (which involves, among other things, a normal perceiver in standard circumstances), agreement among perceivers’ and predictive success-presuppose this worldview of a common space-time receptacle in which the objective accusatives of veridical sense perceptions are the common causes of the, for the most part, nomic-type coherence among the contents of the sense perceptions of the differently positioned perceivers. That there are such tests is not a contingent fact about sense perception but a conceptual fall-out from the concept of a sense perception as having an objective accusative.

These accusatives are ultimately individuated by their position in the space-time receptacle, it being a conceptual truth that numerically distinct individuals of the same kind cannot coincide spatio-temporally. In order to perform this individuating function for empirical entities, these dimensions must not themselves be empirically determined. As Plato said about his Receptacle, ‘that which is to receive in itself all kinds must be free from all characters.’16 This receptacle creates the possibility of there being counter-examples to the principle of the identity of indiscernibles when restricted to fully general properties. Any such property admits of the possibility of multiple instantiations at different regions in this receptacle. As a consequence, we are able to distinguish between perceptions that are of numerically one and the same particular and those of particulars that are only qualitatively similar. In the latter case there are non-coincident particulars that are hooked up with different perceptions via different causal chains.

With this all too brief sketch of the conceptual requirements for sense perception having an objective accusative, the generic requirements for any perceptual experience, including non-sensory religious experiences, having an objective accusative can be derived by abstracting from the features of these requirements that involve or entail having sensory qualities. It cannot be demanded that the objective accusative of religious experiences occupy the spatial dimensions, since being spatial entails having sensory qualities. This requirement, when made properly generic, is that the accusative of a veridical perception of any species occupy some non-empirical dimension(s), though not necessarily those of space and time, by which it is individuated and within which it is causally hooked up with different perceivers, thereby explaining how it is possible both for it to be the common accusative of different perceptions and for the distinction to be drawn between perceptions of one and the same object and those of only qualitatively similar objects.

It is obvious, and readily acknowledged in different things that Swinburne says, that none of these metaphysical requirements, even when divested of all sensory elements as has just been done, are satisfied by religious experiences. To begin with, God is a spiritual being whose individuation is not dimensionally based, not even on time, which he occupies, according to Swinburne. God’s individuation is based instead on his uniquely satisfying certain properties, such as being the sole creator of the world and the determiner of what happens in it, possibly subject to certain limitations due to creaturely free will.

This raises the problem of how we can experientially identify God. To know that one is experiencing God, and not just any old very powerful, loving, non-human person, of which there could be several, it is necessary to know that what one is experiencing completely and solely determines every feature of the world. But this requires knowing the negative fact that there does not exist any being other than this who determines any of these features. The same considerations hold for God’s omni-properties; to know that the apparent object of your experience is omnipotent, you must know the negative fact about it that there is no possible state of affairs that it cannot bring about. Swinburne rightly restricts the PC to positive-seeming experiences: ‘The principle is so phrased that how things seem positively to be is evidence of how they are, but how things seem not to be is not such evidence’ (p. 414; his emphasis). But he fails to see that this restriction to positive-seeming experiences precludes his applying the PC to religious experiences, since they are in part negative. Furthermore, because religious experiences are not dimensionally hooked up to God, there is no way of indexically identifying God. Since we are unable to identify God experientially either indexically or as the unique satisfier of some description, we cannot identify God experientially at all. And this has the consequence that we are unable to determine whether different religious experiences are of one and the same or only qualitatively similar accusatives.

The epistemological requirement. The PC applies to experiences that not only take an objective accusative but also are cognitive in that they count as evidence, even rendering it prima facie probable, that the apparent object of the experience exists and is as it appears to be. But to play this sort of cognitive role, they must be subject to defeaters, tests, or checks, failure of which lowers this prima-facie probability. Roughly speaking, the tests for the veridicality of sense perceptions are being caused in the right way, agreement among observers, and predictability. Since it is reasonably clear how they apply to sense perceptions and how they enable us to distinguish between veridical and unveridical sense perceptions, we can go directly to the question of whether they also apply to religious experiences, bearing in mind that we must drop any feature of these tests that involves sensory qualities.

Being caused in the right way requires a normal observer in a standard situation causally hooked up in ‘the right way’ to the apparent object of the perceptual-type experience. It has already been seen that none of these considerations apply to an experience of God, in virtue of God’s being a non-dimensional (save for time) supernatural being who freely bestows his grace on persons when he causes them to have an experience of him. Thus, the causal test does not apply to religious experiences. And if Swinburne should reply that a religious experience cannot be caused in the wrong way by God, the proper counter is that a test that cannot be failed is no test at all.

Swinburne obviously thinks that religious experiences are subject to the agreement test, for he counts agreement among different observers as confirmatory of the veridicality of their religious experiences (p. 263). But whereas the failure of a sense perception to satisfy the agreement test is taken to be disconfirmatory of its veridicality, Swinburne does not take the failure of a religious experience to pass this test to be disconfirmatory. The reason that failure of agreement is not disconfirmatory for religious experiences but is so for sense perceptions is that only in the latter case do we know what counts as a normal observer and as standard conditions.

If we do not know what experience would count against some perceptual claim (because we do not know which observers could have been expected to have had an experience apparently of s if s had been there), that somewhat lessens the evidential force of an apparent perception-but only somewhat. This is because in that case we cannot have confirming evidence of failure to find evidence which counts against the claim. (p. 175)

The basis of Swinburne’s ‘somewhat lessens the evidential force of’ is obscure. But this slight concession on his part hides the seriousness of the challenge posed to his analogical argument by evidential asymmetry. The in principle impossibility of disconfirming the veridicality of a religious experience by appeal to the testimony of other observers, rather than just lessening the probability of its veridicality, calls into question the very applicability of the agreement test, for it is a strange test indeed that can be passed but not failed. Swinburne seems to have forgotten that he is supposed to be justifying his extension of the PC to religious experience in premiss 2 on the ground that there are no cognitively relevant disanalogies between sense and religious experience. The same considerations hold for his failure to neutralize the disanalogy between them based on the caused-in-the-right-way test being applicable only to sense experiences.

Another example of Swinburne’s failure to face up to a damaging cognitively relevant disanalogy between sense and religious experience is seen in his attempt to neutralize the challenge posed by religious diversity by showing that, for the most part, the statements made on the basis of religious experiences within different religions are compatible, this sometimes, requiring a visit to K-Mart as we saw (pp. 265 ff.). But this overlooks the more serious challenge posed by religious diversity: namely, that whereas all persons who engage in the doxastic practice of making physical object claims on the basis of sense experience agree on what are the relevant tests for veridicality, this blatantly is not the case for all participants in the doxastic practices of basing claims about God on religious experiences: the tests for veridicality employed within the different religious traditions are mutually incompatible, as, for example, in what they count as growing in sanctity and the revealed truths that are recognized as normative.

It is to William Alston’s credit that, in Perceiving God, he recognized that this is the real challenge posed by religious diversity. Unfortunately, his attempt to counter it by arguing that since the rival religious traditions do not share a common method for warranting or overriding claims based on religious experiences, the epistemic discreditation is far less than it would be if they shared the same method but differed in their output beliefs, fails. Certainly persons holding rival beliefs would be more seriously divided if they could not agree on a method for resolving their disagreement than if they did; for in the latter case they can at least argue with each other, unlike the former, in which they can resort only to nonepistemic means to resolve their differences. Furthermore, someone impugns my epistemic soundness far more if they question the very method by which I arrive at and warrant my beliefs than if they only question some of my beliefs.

There still remains the prediction test. There are two issues here: being able to predict who will have the type of perceptual experience in question and being able to predict future events that will confirm the veridicality of a given experience of that type. Both can be done for sense, but not for religious, experience. Because a religious experience isfreely brought about by God, there is no way to predict who will have such an experience and under what conditions. It is for this reason that Swinburne calls religious experiences ‘private’ (p. 269). The best that he can offer in response is: ‘if there is a God, there is a greater probability that men will have such experiences’ (p. 275); and it is more likely that the person of faith, rather than the atheist, will have them, for ‘there are no grounds for supposing that if there is a God, the atheist would have experience of him’ (p. 310).

Both of these extremely vague probabilistic predictions can be challenged. A religious experience is often so psychologically overpowering that the subject is forced to believe that God exists, which goes counter to Swinburne’s requirement that such belief be free. Such an experience would not exert as coercive a force upon an atheist, since it would be counterbalanced by his initial disbelief. The atheist, furthermore, being far more in need of God’s grace than the person who is already launched along the path of faith, would seem to be more likely to have a religious experience if God exists than would the believer, pace Swinburne. I am not, of course, altogether sincere in these probabilistic pronouncements, but merely wish to show how little basis there is for making any probabilistic predictions at all in this area.

As for being able to predict future events that would be confirmatory, Swinburne appeals to the traditional test based upon the subject’s growth in sanctity (p. 273). Given God’s omnibenevolence, we would expect that a veridical experience of him would result in the subject becoming more God-like. Not only does this encounter the problem posed by religious diversity, due to there being incompatible criteria for growth in sanctity within different religions, it also strikes out on Bayesian grounds. The prior probability that S has a veridical religious experience (p) relative to background knowledge k is very low, and p‘s explanatory or predictive power in conjunction with k of the fact that S becomes more sanctified (e) is no greater than that of k alone in respect of e-P(e/p.k) = P(e/k). The reason for this is that k includes known facts about the psychological causes of human behaviour that explain why someone who believes that she veridically experiences God is likely to become more holy. Thus, on Bayesian grounds, P(p/e.k) is very low.

In conclusion, religious experiences not only do not have any defeaters analogous to those for sense experience, they have no defeaters at all. For this reason alone, they fail the epistemological requirement for being perceptual-type experiences. And when this failure is combined with their failure to pass the metaphysical requirement, there is more than ample justification for refusing to extend the PC to them. We have thereby met Swinburne’s challenge to produce a relevant disanalogy between sense and religious experience that would justify applying the PC to the former but not to the latter.


  1. Oxford: Clarendon Press, 1979. References to this book are included in the body of the text.
  2. Faith and Reason (Oxford: Clarendon Press, 1981), 5-6.
  3. The very use of Bayes’s theorem is suspect. According to this theorem: P(h/e.k) = P(e/h.k)/P(e/k) X P(h/k)

    Since a numerical value can be assigned neither to h‘s prior probability-P(h/k)-nor to h‘s explanatory power-P(e/h.k)/P(e/k)-it is unclear what it means to multiply them together.

  4. I think that Swinburne errs in making both recognitional and inferential perceivings of God subject to the PC. In regard to seeing a sunset as an expression of God’s love he writes: ‘That God is at work is no inference for these men but what seems… to be happening’ (p. 253). People in our human epistemic circumstance who say that they perceive a sunset as an expression of God’s love, unlike those who say that they simply feel God’s presence or God’s comforting them, are making inferences to the existence of God as witnessed by the fact that they are willing to give inductive backing to their assertion, such as is supplied by Swinburne’s cumulative P-inductive argument.
  5. This definition is too broad if we take every non-spatio-temporal entity to be supernatural; e.g., we would not want to call Plato’s intuitive experience of triangularity a religious experience.
  6. William P. Perceiving God: The Epistemology of Religious Experience (Ithaca, NY: Cornell University Press, 1991), 195.
  7. In regard to the latter, consider Nouet’s description of a religious experience, quoted in Anton Poulain, The Graces of Interior Prayer, trans. L. Y. Smith (London: Kegan Paul, Trench. Trubner, 1950), 111: ‘God, who was formerly in the soul of the just as a hidden treasure, by way of sanctifying grace, now presents Himself to her as a Treasure that is found. He flows into her faculties. He gives Himself to her. He fills her with the fullness of His Being. The soul, in return, ravished by His charms and by the spectacle of His beauty, holds Him, embraces Him, clasps Him closely, and all on fire with love, she flows, she plunges, she buries and loses herself deliciously in God with sentiments of inconceivable joy.’ Why, this is nothing but smut, and we should take steps to have Poulain’s book banned from our schools.
  8. I failed to see this possibility in my book On the Nature and Existence of God (Cambridge: Cambridge University Press, 1991), 315, when I said that ‘it is impossible for God to cause an of-God experience in the wrong way’. I am thankful that there are no malpractice suits in philosophy, or I would have been seduced to the state of pauper a long time ago.
  9. Gary Guting, Religious Belief and Religious Skepticism (Notre Dame, Ind.: University of Notre Dame Press, 1981), 149.
  10. I have corrected a typo in the text: an ‘S’ incorrectly appears in place of my final ‘x‘.
  11. Swinburne claims that the probability that the apparent experience of x is veridical is increased if either the experience is forceful (p. 265) or others have similar experiences (p. 263). Forcefulness would seem to have nothing to do with the case, especially since many unveridical experiences are very forceful, and the vast majority of our waking veridical experiences are quite unforceful, pace Hume. Furthermore, it is highly suspect to allow for confirmatory experiences of the presence of x when the possibility of disconfirmatory experience is ruled out.
  12. Another place at which this failure becomes evident is in his attempt to deduce God’s omnibenevolence from his omniscience and perfect freedom, on the assumption that there are objective moral truths (p. 101). What this Socratic all-virtue-is-knowledge approach overlooks is that someone could know what is the morally right thing to do and still perversely choose the immoral alternative.
  13. For an excellent discussion of this issue which supplies the technical details missing in my account, see Robert M. Adams, ‘Plantinga on the Problem of Evil’, in I. Tomberlin and P. van Inwagen (eds.), Plantinga (Dordrecht: Reidel, 1985), 225-55.
  14. For a fuller discussion of this, see my article, ‘Why Alston’s Mystical Doxastic Practice is Subjective’, in the Philosophy and Phenomenological Research symposium book on William Alston’s Perceiving God (forthcoming).
  15. For a fuller account of this, see ch. 8 of my book On the Nature and Existence of God.
  16. Plato, Timaeus 50E-51; quoted from F. M. Cornford, Plato’s Cosmology (New York: Liberal Arts Press, 1952). This issue is dealt with more fully in my book Negation and Non-Being, American Philosophical Quarterly Monograph no. 10, 1976.