Home » Library » Modern Library » Science, Confirmation, and the Theistic Hypothesis: Chapter 2

Science, Confirmation, and the Theistic Hypothesis: Chapter 2

Chapter Two

Schlesinger on the Confirmation of Theism

The first lesson to be learned from the failure of “scientific” creationism is that theism should not be presented in the guise of a scientific theory. Any such effort is bound to lapse into pseudoscience in just the way and for the same reasons as did creationism. As we saw in the last chapter, a theory, T, will count as scientific only if it is capable of being tested vis-a-vis rival theories. This means that T must be capable of (a) at least matching the explanatory power of its rival theories and (b) clashing empirically with its rivals by generating test implications that would be very probably false if the rival theories are true. With respect to (a), it is hard to see how theism can explain phenomena otherwise than as the results of the volitions of a deity. As we saw in the last chapter, explanations of this sort fail to subsume individual events under general nomologicals. Hence, they are unrepeatable, untestable, and cannot be classed as scientific explanations. Further, if it is held that such explanations do not exclude, or at least make highly improbable, any empirical circumstance, there can be no way for theism to meet condition (b).

Given, therefore, that theism is not, as science is presently conceived, a scientific hypothesis, it must be further insisted that theological tenets should never directly conflict with scientific beliefs. Whenever theists ignore this rule, as did Fundamentalists at the Scopes trial, they inevitably cast themselves in the roles of obscurantists. For example, those who insist on the occurrence of the Noachian flood in the face of overwhelming geological evidence to the contrary, cannot avoid the appearance of intransigence and dogmatism. However, even fundamentalists now realize that it is unwise to place nonscientific claims, e.g. those based entirely upon a purported revelation, in direct competition with scientific ones. This is why they are at such pains to emphasize the “scientific” character of creationism.

A corollary to the rule that theism ought never to oppose science is that all “God of the gaps” arguments must be eschewed. In other words, God should not be invoked to account for phenomena that science cannot presently explain but might well come to understand in the future. Such efforts to erect theism in the explanatory niches left blank by contemparary science can only result in embarrassed retreat as these lacunae are progressively closed.

Some philosophers, while conceding that theism is not a scientific theory or any sort of competitor to science, maintain that the employment of certain of the principles of confirmation used in the evaluation of scientific hypotheses can nevertheless lend strong support to theism. The two best known such philosophers are George Schlesinger and Richard Swinburne, each of whom has written extensively on confirmation theory.[1] Their arguments comprise an effort to recast some of the old theistic “proofs” in probabilistic form. That is, while they do not think that the existence of God can be made demonstratively certain, they hold that the tools provided by confirmation theory allow theism to be established as a very likely hypothesis.[2]

This chapter will examine Schlesinger’s attempt to produce inductive teleological arguments and the next chapter will take up Swinburne’s similar efforts with respect to the cosmological argument. I shall show that neither succeeds in showing theism to be a likely hypothesis or even to be more likely than many competing hypotheses. Although Schlesinger and Swinburne share a number of points, it will be more convenient to examine their arguments separately. Since Schlesinger’s case is presented more straightforwardly, I shall examine his first.


If theism ought not to be considered a scientific hypothesis, what kind of hypothesis can it be? Schlesinger’s answer is that it is a metaphysical hypothesis.[3] Metaphysical hypotheses are like scientific ones in that they both attempt to account for the data of experience. However, there is an essential difference in the way that empirical evidence relates to the two sorts of hypotheses. When two scientific hypotheses clash, the accumulation of evidence will eventually lead to the rejection of one of them. That is, continuing observation and experiment will produce results that fit smoothly with one hypothesis but can be made consistent with the other only through the introduction of ad hoc auxiliary hypotheses. Finally a point will be reached when one hypothesis will be so encumbered with ad hoc additions that it will be rejected on grounds of elegance or simplicity in favour of its rival.[4]

On the other hand, conflicts between metaphysical hypotheses are not resolvable in this way. It is, for instance, very difficult to see how the debate between materialism and idealism could ever be solved through the accumulation of empirical evidence. Similarly, no amount of empirical evidence would seem to count in the debate between realists and nominalists or in the McTaggart-Russell controversy over the nature of time.[5] In other words, metaphysical hypotheses generate no test statements that would permit future observations or experiments to decide between them. Schlesinger concedes that unanticipated discoveries could lead to the resolution of issues that had been thought metaphysical. However, he contends that the discovery of such empirical evidence would, ipso facto, show that those issues were not in fact metaphysical.[6]

Schlesinger follows tradition in regarding theism as a metaphysical hypothesis and so regards its relationship to the empirical situation as essentially static. Thus, he holds that there are no arguments presently available to either theist or atheist that could not, in principle, have been advanced 300 years ago.[7] Further, if theism is indeed a metaphysical hypothesis, no new observations can arise that will count either for it or against it (the obvious counterexample to this claim–miraculous occurrences–will be the topic of a later chapter).[8]

Given this view, it is not surprising that Schlesinger should be worried about Flew’s falsificationist challenge, understood by him as follows:

Flew requires–and it does not seem unreasonable to require–that statements which purport to be statements describing how the world is conform to the verifiability principle to which all statements, the meaningfulness of which is to be established, conform. The version of the principle to which theological statements and all other such statements have to comply is… “A sentence is confirmable if, and only if, circumstances are describable under which it would be regarded as confirmed on the basis of prevailing tenets of scientific methodology.”[9]

But if no future discoveries could count either for or against theism–in short, if it implies no test statements–how can it possibly be confirmable?

Schlesinger holds that hypotheses may be confirmable even if they are not directly testable. He argues that some of the basic principles of confirmation theory apply to all hypotheses, regardless of whether they stand in a static or a dynamic relationship with the empirical evidence. He believes that two of these principles are adequate to show that theism is more highly confirmed than any alternative hypothesis. His first principle is as follows:

Principle E: when a given piece of evidence E is more probable on H than on H’ then E confirms H more than H’.[10]

He holds that E can be quite easily justified:

This I believe should appear very reasonable to everyone and can be shown to be so in various ways. After all, when the probability of E on H is zero, that is when p(E/H)=0, then E falsifies H, thus the further p(E/H) is from zero the less E falsifies H or the more it confirms it. Or, one may say, the hypothesis we wish to adopt is to account for E, and it is clear, a hypothesis on which E is highly improbable does not account for E; in fact the very meaning of the expression “account for” implies that the more an hypothesis makes E probable the more it may be said to account for it.[11]

As an illustration of the working of Principle E, we may consider two hypotheses: “All ravens are black” and “Ravens come in more than one colour.” If the evidence we are attempting to account for is “Of the five ravens hitherto observed have been black,” then it is clear that this evidence is much more likely if the first hypothesis is true than if the second is.[12] Hence, we may conclude, through the application of E, that if all the ravens hitherto observed have been black, this confirms the hypothesis that all ravens are black more than it confirms the hypothesis that they come in more than one colour.

Suppose, however, that the evidence is equiprobable with respect to two hypotheses, that is, that p(e/h) is not appreciably different from p(e/h’). It might nevertheless be that h is to be preferred over h’ because h is intrinsically more “adequate” than h’. Schlesinger expresses this through his Principle A:

Principle A: When H and H’ are similarly related to all the available evidence, we regard H as more confirmed than H’, if and only if, H is more adequate than H’.[13]

But what does Schlesinger mean by “adequate?” He answers by first pointing out (as did Rothbart in the last chapter) that with respect to any piece of evidence there are infinitely many hypotheses that could equally well account for that evidence. Schlesinger claims that the only guiding rule that allows us to select a single, specific hypothesis from the infinity of contenders is the principle of parsimony. That is, he argues that any guiding rule that instructed us to pick other then the simplest available hypothesis would still leave us with an infinity of hypotheses to choose from. Hence, if the principle of parsimony is the only usable guiding rule, “adequacy” must be defined in terms of that rule: “the most adequate hypothesis is the one selected with the aid of the only useable guiding rule.”[14]

As an example of the application of Principle A, suppose we had made many observations of the distances covered by falling bodies near the surface of the earth. The simplest formula correlating the distance with the duration of such falls is the one advanced by Galileo: s=1/2gt2 (where s is the distance covered, g is a constant, and t is the duration of the fall). However, all of our observations equally accord with an infinite number of alternative hypotheses that have the form s=1/2gt2 + f(x), where f(x) is a function equalling zero for all observations made so far but perhaps not equalling zero in the future. Hence, the only way to pick a specific hypothesis from this infinity of candidates is to follow the rule of parsimony by selecting Galileo’s formula.[15]

However, even if Principle A is the only useable guide for the selection of hypotheses, what justifies our further belief that the hypotheses picked out by A are in fact true? That is, even if practical exigencies force us to produce some hypothesis, and hence to employ A, what justifies our faith that the hypotheses so selected depict the way things are? Schlesinger gives an answer along the lines developed by Strawson in his defense of induction.[16] Part of what we mean when we claim that our beliefs are reasonable and rational is that we hold them if and only if they were arrived at by methods that are themselves reasonable and rational. If A is the only useable guide for the selection of hypotheses that are symmetrically related to the evidence, then A would seem to be the only reasonable or rational rule to employ. Hence, by definition of “reasonable” and “rational,” it follows that an hypothesis selected through the application of A must itself be reasonable and rational and thus that confidence should be placed in its predictions.[17]

Now that he has articulated and justified his two fundamental Principles, Schlesinger is prepared to employ them with respect to the theistic hypothesis. Schlesinger begins with premises that he considers unquestionably true:

… it is unquestionably true that human beings exist who possess awareness and conscience and are capable of a wide range of emotions and sentiments. It is also undisputably true that the laws of nature which govern the universe, and the initial conditions, are such that complex and precarious systems like humans can come into existence and survive. Human beings are creatures who are capable of responding to the Divine. By this I mean that they have the potential for acknowledging and contemplating God, as well as having feelings of awe and love toward him.[18]

Schlesinger holds that the laws of nature and the initial conditions that allowed the development of creatures capable of responding to the divine are facts that need to be accounted for. Let R stand for those laws of nature and initial conditions. Schlesinger contends that the theistic hypothesis accounts for R much better than naturalism. Given the hypothesis than an omnipotent being exists who desires that creatures come into being who are capable of responding to the divine, it is certain that such creatures will exist. In other words, if T stands for the hypothesis that an omnipotent being exists who wills the existence of beings capable of responding to the divine, p(R/T)=1.

On the other hand, if N stands for naturalism, i.e. the denial that any supernatural power exists behind nature, then p(R/N) < 1. This is because naturalism is compatible with the actualization of any possible world, including the many worlds in which creatures capable of responding to the divine would never exist. Thus p(R/T) > p(R/N). Since, therefore, R is more probable on T than on N, Principle E tells us that R confirms T more than it does N. In short, the existence of creatures capable of responding to the divine makes theism more probable than naturalism.[19]

Curiously, Schlesinger follows his more recent statement of the above argument with a set of apparently devastating criticisms. Three of these criticisms are sufficiently important to be summarized in some detail:[20]

(1) In all of our usual assessments of probabilities–from the rolling of dice to the behavior of quanta–we presuppose some empirical basis for such judgments. However, we have no experience whatever of unactualized universes and hence no grounds for assigning degrees of probability for the actualization of such possible worlds.

(2) We have no right to assume that each possible world has an equal probability of being actualized. True, there are infinitely many worlds in which creatures capable of responding to the divine never exist, but there are also infinitely many worlds in which they do. Hence, without more information on how to compare the relevant sets, we cannot be sure that p(R/N) < 1.

(3) Ordinarily, as when we are assessing the probability of a certain roll of a die or the decay of a particular atom, we assume that the events we are passing judgment on are physical processes. In fact, it seems that meaningful probability judgments can only be made with respect to well-defined physical processes. However, there seems to be no conceivable physical process whereby possible worlds could turn into actual ones.

Given the above difficulties in establishing probabilities for the actualization of possible worlds, Schlesinger develops a new argument that makes reference only to the actual world.[21] In order for humans to be capable of responding to the divine, they must be free moral agents who can plan and carry out purposeful actions. Agents can perform purposeful acts only if they are capable of knowing what the results of their acts are likely to be. For example, virtuous acts would be impossible (at least on a consequentialist understanding of the nature of virtue, which Schlesinger apparently presumes) unless they were known to be very likely to bring about some specific good rather than just any haphazard result. Let V=My virtuous act A, designed to bring about some good result T, is going to succeed in producing T.[22] Clearly, in order to perform A, I must know that V is very likely the case, i.e. that my virtuous acts are very likely to bring about their intended results. Further, it seems that the only way that V could be known is through inductive generalization based on past experiences of the results of A-like actions.

Schlesinger argues that p(V/N) is considerably less than one. There are three reasons for this:

1 The premises of every inductive argument contain assumptions concerning the prevailing physical conditions. We can never be certain of the truth of these assumptions.

2 Even if the inductive method has been proved absolutely reliable, we could never be sure that we have applied it correctly, that is, that we have made all the relevant observations and have an adequate sample class.

3 The greatest source of uncertainty is our continued inability to show that events of the past are capable of providing any support for hypotheses about the future. Many philosophers are pessimistic about the prospects of attempts to justify induction. But even among those who have offered such justification, most have not claimed to have provided reasons for regarding the conclusions of inductive arguments as highly probable.[23]

For the theist, however, in addition to any arguments that the naturalist might give in justification of induction, there is an additional reason for thinking that V is true.[24] Since God has created creatures whom he expects to act virtuously, he will guarantee that V is the case. Hence, p(V/T) > p(V/N). Principle E says that if something is more likely on hypothesis A than on hypothesis B, its actual occurrence confirms A more highly than B. Thus, Schlesinger claims that it is more likely that the probable consequences of virtuous actions will be known if God exists than if he does not exist. Hence, the fact that V is known to be true, i.e. that acts designed to bring about good results are known to be likely to do so, shows that theism is more likely to be true than naturalism.

However, even if it is conceded that all the relevant evidence E confirms T more highly than N, might there still not be some hypothesis D, incompatible with T, such that p(E/D)=p(E/T)?[25] For instance, might not some evil demon have created the world and endowed it with all of its present features in order to serve some diabolical scheme? Schlesinger holds that his Principle A can be employed to pick T out and raise it above all the D-like hypotheses.

Principle A tells us that when two hypotheses are symmetrically related to the evidence we are to pick the more adequate one. The more adequate one is the one selected by the only useable guiding rule, i.e., the only rule that enables us to select a single, determinate hypothesis. With respect to the choice between T and all the D-like hypotheses, Schlesinger maintains that the only useable guiding rule is the one that enjoins us to select the hypothesis postulating a perfect being as creator.[26] Precisely how this guiding rule relates to his earlier recommendation of the principle of parsimomy is not made clear. Perhaps Schlesinger holds that the hypothesis postulating a perfect being will ipso facto be the most parsimonious and so this latter rule is merely a rewording of the earlier one in a way that makes it more applicable to this context.

At any rate, Schlesinger defends his injunction to select the perfect being by arguing that perfection is the only attribute that cannot be shared by non-identical individuals. Clearly, two non-identical individuals can have the same color, shape, size, etc., and hence these properties cannot be used to pick out a unique individual. Only the property of absolute perfection picks out a unique individual–an individual identical with the being postulated by theism.

Why, though, cannot there be more than one perfect being? What is there about perfection, and perfection alone of all possible attributes, that necessarily makes it possessor unique? Schlesinger does not answer this directly, but he adumbrates a reply when he claims that if we know that a being is perfect we will immediately know what other properties that being must possess. With any imperfect being, on the other hand, we will be left in the dark about many predicates:

Anyone who wants to postulate that the being assumed responsible for everything that exists is not perfect is obliged to make indefinitely many arbitrary decisions about whether this or that property should be assigned to this being independently of all the other properties already assigned to him. After having ascribed n properties to the powerful ruler of the universe, no matter how large n may be, the question will always remain whether some property P, not included among these and unrelated to them, belongs or does not belong to him. The theist, on the other hand, would claim that has to do is ask himself whether P or not-P contributes more to perfection, a question he would insist always has a clear answer.[27]

The implications of this argument need to be spelled out in somewhat greater detail. In order to do so, Schlesinger’s somewhat meagre remarks will have to be amplified in a way that is intended to capture the full force of his arguments: With respect to any hypothesized imperfect being B, no matter how many properties we have ascribed to B, there will be an indefinite number of other properties which B could coherently be supposed to either possess or not possess. Hence, there could be some other being C that is identical to B with respect to all the properties we have so far considered but which differs from B with respect to the possession of some property not yet taken into consideration. Thus, no matter how many properties we ascribe to an imperfect being, we will not have succeeded in picking out a unique individual. For Schlesinger this further implies that, since an indefinitely large number of conceivable beings could be the hypothetical imperfect creator, there can be no single, determinate hypothesis postulating such a being.

With an absolutely perfect being, however, there will be precisely one set of properties that that being can possess. A perfect being will possess all of those properties that contribute to perfection and lack all of those properties that detract from perfection. Further, it is assumed that no properties are neutral in the sense that it is irrelevant to a subject’s perfection whether those properties are present or absent in that subject. Hence, since the property of perfection determines all other properties, there cannot be two non-identical perfect beings. A quick reductio ad absurdum shows why this must be so: Suppose there were two perfect beings, P and P*. Since perfection determines all other predicates, P and P* must be exactly alike in every respect. But if P and P* are exactly alike in every respect they will be identical, which contradicts our supposition. It therefore appears that the criterion of perfection is the only one that permits selection of a unique being from among the infinity of hypothetical creators. Further, the hypothesis postulating such a being will be the only single, determinate one.


In assessing Schlesinger’s arguments, it will initially be presumed that his Principles A and E are sound. I shall claim that even if A and E are allowed to stand as established principles of confirmation theory, the theistic arguments he bases on them are nevertheless unsound. It will then be seen that Principles A and E are themselves faulty and so cannot be employed to produce viable theistic arguments.

The first thing to note is that, despite his claim to have done so, Schlesinger has not met Flew’s challenge. Flew’s specific demand is to know what would count against theistic claims.[28] That is, he wants to know whether the theist would count any conceivable circumstance as incompatible with such putative claims as that God created the world or that he loves his creatures. The beauty and power of Flew’s challange lies in the fact that it is based, not–as many have wrongly supposed–on a dogmatic and now thoroughly discredited positivistic criterion of meaningfulness, but on one of the simplest principles of logic. That principle is that as assertion is equivalent to the denial of its negation: P is equivalent to ~~P. Hence, “God created the world” is necessarily equivalent to “It is not the case that God did not create the world.” Now the two preceding statements (in reality different wordings of the same statement) have every appearance of being, and have historically been regarded by those who asserted them as being, factual assertions. But if no conceivable evidence can be taken as an indication that God did not create the world, the claim that he did not, and consequently the denial that he did not, will be, if not meaningless, at least entirely pointless and vacuous.

This is not the place to give a detailed defense of Flew’s challenge, especially since many others, including Schlesinger, have done so with considerable cogency.[29] Here it is sufficient to stress that Schlesinger accepts that challenge and explicitly claims to have met it by showing that theism is in fact confirmed.[30] However, though he tells us that R and V serve to confirm T, Schlesinger gives no indication of the sort of evidence that would serve to disconfirm that hypothesis. In fact, he devotes considerable energy to showing why the evils and imperfections of the cosmos cannot constitute such negative evidence.[31]

More fundamentally, since Schlesinger does not hold that theism is presently falsified, he can meet Flew’s challenge only by claiming that some conceivable yet presently unavailable evidence would indeed falsify theism. Yet if theism is a metaphysical hypothesis in the sense that Schlesinger claims it to be, all the evidence must be in and no future evidence could add to its confirmation or disconfirmation. Hence, on Schlesinger’s construal there seems to be no way for theism to meet Flew’s challenge–unless, perhaps, it is already falsified.

Despite Schlesinger’s expertise in confirmation theory, one critic notes that he makes a rather elementary mistake.[32] Schlesinger conflates two distinct types of inductive arguments–arguments that his colleague Swinburne distinguishes as P-inductive and C-inductive.[33] A good P-inductive argument, is one that succeeds in showing that its conclusion is probably true. A good C-inductive argument is one in which the premises make the conclusion more probable than it otherwise would be, but which do not necessarily show that the conclusion is overall more probable than not. Hence, even if Schlesinger’s arguments are good C-inductive arguments, it does not follow that they succeed in making T overall more probable than N. In other words, it might be that R and V each confirm T more highly than N, but on the basis of the complete available evidence N might be better confirmed than T.

Further, Schlesinger admits that Principle E does not automatically make the overall probability of one hypothesis higher than another.[34] For instance, it might be that a given piece of evidence e confirms hypothesis h more highly than hypothesis h*. However, it might be true that the antecedent probability of h is so much lower than that of h* that h* is still the more probable of the two. Some atheists have in fact argued that T is antecedently far less probable than N.[35] Thus, it would have been more satisfactory if Schlesinger had preceded his arguments with a rebuttal of such atheistic claims (we shall see that Swinburne does address this issue).

With respect to his first argument, that R confirms T more highly than N (hereafter this argument will be referred to as RA), Schlesinger’s own criticisms seem to be quite devastating. Clearly, if there is no way to determine p(R/N) there can be no grounds for asserting that p(R/T) > p(R/N). However, even if some means could be achieved that would permit the assignment of probabilities for the actualization of possible worlds, RA still might face a weighty difficulty. That difficulty is that some physicists, advocates of what is called the “strong anthropic principle,” do hold that at least one such probability can be assigned.[36] These physicists contend, apparently basing their arguments on nothing more than tautological background knowledge (e.g. the laws of logic and mathematics), that the universe had to produce conscious observers. If true, this might land strong support to the claim, incompatible with RA, that P(R/N)=1.

However, the anthropic principle is a highly controversial one and is by no means acceptable to all physicists. The fundamental difficulty with RA is that Schlesinger radically misconstrues naturalism by interpreting it as an explanatory hypothesis. That is, Schlesinger views naturalism as a hypothesis from which certain probabilities can be derived in order that they may be compared with probabilities based on a rival hypothesis, viz. theism. But it is highly questionable whether naturalism qualifies as any sort of hypothesis.

It will be recalled that Schlesinger defined naturalism purely negatively as the denial that any supernatural power exists behind nature.[37] But even this is putting it a bit strongly. Naturalists do not have to deny outright that supernatural powers exist; they need only to deny the need for any explanation going beyond or outside the universe.[38] That is, naturalists hold that, if explanation must ultimately reach a terminus (an assumption that theists must make), there seems to be no reason why that terminus cannot be the most fundamental features (whatever they may be) of the physical cosmos itself. So defined, naturalism makes no positive claims about the universe. It does not identify any feature of the universe and claim it to be the ultimate existent, the terminus of explanation. It merely holds that if there is a fundamental physical reality, there is no reason why it cannot constitute such an ultimate terminus. Hence, naturalism is not an hypothesis; it is the denial of the need for certain kinds of hypotheses. In fact, naturalists see no reason to even regard the natural/supernatural dichotomy as a legitimate ontological distinction.

This denial of the need for hypotheses positing supernatural entities must, of course, be accompanied by the claim that no one has in fact succeeded in showing the truth of any hypothesis that invokes cosmos-transcending explanations. However, this is a claim about arguments, not about the world. In fact, even this claim need not be based on any positive evidence (though, as we shall see in the chapter on the problem of evil, such evidence exists). It can arise from the belief that no arguments have yet succeeded in overcoming what Antony Flew has called the Stratonician Presumption (after Strato of Lampsachus, head of Lyceum from about 286-269 B.C.): “The claim that it is up to anyone wanting to postulate a God (or any other initiating or sustaining principle outside and beyond the Universe) to show sufficient reason for doing so.”[39]

The Stratonician Presumption is therefore a methodological prescription: Let the burden of proof fall on those who want to invoke explanations that transcend the physical cosmos. The justification for making this Presumption is simply that theism is necessarily a less parsimonious view than naturalism. The theist believes in the existence of all that the naturalist believes in–the physical cosmos and the features that science currently holds it to possess–but he/she then goes on to postulate the existence of one additional entity. If our beliefs ought to be justified (and this seems to be a requirement of every conception of rationality), then the theist has one more belief to justify than the naturalist.[40]

Naturalists are those who make the Stratonician Presumption (hereafter abbreviated as SP) and further hold that no one has yet succeeded (or, more radically, can succeed) in meeting the burden of proof it imposes. Naturalism, since it is not an hypothesis, has no bearing on the likelihood of the existence of the universe or any of its features. For naturalism, the universe is neither likely nor unlikely, probable nor improbable; it simply is. It follows that Schlesinger’s Principle E cannot be employed to show that p(R/T) > p(R/N). The right hand side of this inequality, p(R/N), since there is no hypothesis N, has no value either high or low. In consequence, RA, since it wrongly assumes that naturalism is a hypothesis that can be compared with theism on that basis of Principle E, is not, as it stands, a good argument.

If Schlesinger is to make any headway against the naturalist he must overcome the SP. That is, he must present the naturalist with some reason to think that the universe or some feature of it, such as its possession of R, needs to be, explained. Schlesinger might reply that he has accomplished this when he argues that, when we consider the infinite number of possible universes, the chances of getting one like ours is infinitessimal. Here the naturalist need only reiterate the points, made by Schlesinger himself, about the difficulties surrounding the attempt to establish probabilities for the actualization of possible worlds.

Schlesinger’s argument based on V, the claim that the likely consequences of acts intended to bring about good results can be known, will hereafter be abbreviated as VA. VA follows RA in wrongly assuming that naturalism is a hypothesis. However, it must inevitably appear cavalier to simply dismiss the argument on that basis. Hence, this and the remainder of Schlesinger’s arguments will be criticized on their own terms, i.e. with the assumption being made that naturalism is an hypothesis.

First, whereas RA at least begins with a plausible premise, that there exist creatures capable of responding to the divine, VA does not. In fact, skeptics such as Hume would explicitly deny that they know that V is true. Hume would argue that in order to act virtuously it is not necessary that we have sound inductive knowledge of the consequences of virtuous acts. True, experience imbues us with the habitual expectation that certain consequences will follow from certain acts, and this is all that is required in order to act virtuously. However, such habits do not for Hume constitute knowledge.

Nevertheless, it is undeniably the case that the great preponderance of people would be prepared to say that we have solid inductive grounds for holding V to be true. There do seem to be certain acts that we know are very likely to bring about the good results they are intended to achieve. However, there is no reason to hold that V is more probable on T than on N. The first two reasons given by Schlesinger for holding that p(V/N) < 1 apply just as much to T as they do to N. There is simply no reason to think that a theist is any less likely than a naturalist to be mistaken about the prevailing physical conditions that are assumed in the premises of an inductive argument. Neither is the theist any more infallible when it comes to making the correct application of inductive methods.

As for the claim that the naturalist ought to have less confidence in induction than the theist, the best rebuttal issues from Schlesinger’s own writings.[41] According to Schlesinger, belief in the validity of induction is a metaphysical belief. Further, he claims that it is characteristic of some metaphysical beliefs, including the belief in induction, that it is perfectly rational to hold them with complete assurance even when they cannot be justified:

Metaphysical facts of the kind we have considered … are so conspicuous and inescapable that we intuit them directly without going through any process of derivation. Consequently a belief in certain true metaphysical propositions is self-certifying and a person may acquire it with complete assurance. It might be asked why, if this is so, we have such an extensive literature of arguments designed to establish metaphysical propositions. The answer would be that these are explanatory arguments intending to show why these propositions are true and not that they are true … For example, from the time we began to reason we all took it for granted that induction is valid and never entertained any real doubt about it [emphasis in original].[42]

Hence, the naturalist would only seem at a disadvantage to the theist when it comes to saying why confidence should be placed in induction. They are on an equal footing when it comes to asserting that induction is reliable. Hence, the belief that V is true seems to be just as well grounded on N as on T.

Finally, it seems plausible to argue that if a God existed who truly wished us to be virtuous, He would have given us much better knowledge of the consequences of our actions than we presently possess. There are many situations in which something demands to be done but we are at a loss about what to do because, due to our ignorance of the probable consequences of our actions, it is unclear which act is virtuous. Hence, our vast ignorance of many of the consequences of our acts seems to constitute prima facie evidence for N over T.

If we now suppose, despite the failure of RA and VA, that T is overall more probable than N, there is still the problem of alternative hypotheses, i.e. that there might be D-like hypotheses on a par with T with respect to the evidence. This remains a problem even if the soundness of Principle A is conceded and the injunction to select the perfect being is obeyed. Contrary to what Schlesinger claims, it does not always seem clear whether certain attributes are implied by perfection or not. For instance, does perfection entail that God exists as three persons in one substance or as a single person? Christians would give one answer and Muslims another though both would agree that God is to be conceived as perfect. How do we decide which is right (if, indeed, either is)? Until this question is answered there will be at least two hypothetical entities that the phrase “the perfect being” could be taken as referring to. In this case, the injunction to select the perfect being will not have succeeded in narrowing the field of candidates to a single being and so cannot meet Schlesinger’s adequacy requirement.

Supposing, however, that Schlesinger could provide arguments that would settle such questions to the satisfaction of all the qualified, rational parties (not a very likely prospect given the history of theological dispute), there would remain a fundamental wrongheadedness about his whole approach. Specific, determinate hypotheses can be produced without invoking unique beings. The hypothesis that the extinction of the dinosaurs was caused by the collision of an asteroid with the earth is a determinate hypothesis. This is so despite the fact that any one of an indefinitely large number of conceivable asteroids could have been the one that supposedly brought about the dinosaurs’ demise. Likewise, the hypothesis that humans and the great apes had a common ancestor is a distinct, determinate hypotheisis even though there are many features that such a hypothetical ancestor could have had or lacked. Thus, by analogy it would seem that the hypothesis that the universe is the creation of a powerful but imperfect being with properties P-1, P-2, P-3, … P-n is a determinate hypothesis despite the fact that an indefinitely large number of beings could conceivably meet this description.

As to what properties P-1, P-2, P-3, … P-n are to be assigned, they will be the ones that best fit the available evidence and also meet the other accepted criteria for hypothesis selection (e.g., parsimony, fecundity, concordance with accepted theories, etc.). With respect to this last point, the hypothesis of a limited and imperfect creator seems prima facie a much more modest and parsimonious hypothesis than one postulating a perfect being. That is, when the evidence to be explained is our finite and imperfect cosmos, it seems unnecessarily extravagant to invoke a perfect being when an imperfect one would do just as well. In fact, it might be argued that an imperfect one would do better. As Hume’s Philo contended, the world can plausibly be seen as more like the work of a third-rate apprentice than a perfect creator.

So far it has been assumed that Schlesinger’s Principles A and E are unobjectionable; only the arguments based on them have been criticized. However, as they stand, i.e. without extensive qualification, they do not constitute sound principles of confirmation.

Unless it is greatly supplemented or modified, Principle E can be employed to “confirm” hypotheses that scientific reasoning is bound to find unacceptable. For instance, let

h=All ravens are black because the Devil made them black.

h*=99% of all ravens are black.

E=Of the five ravens hitherto observed all were black.[43]

Now, since p(e/h) > p(e/h*), Principle E tells us that e confirms h more highly than h*. Yet something has surely gone wrong when a purportedly scientific procedure confirms hypotheses such as h more highly that hypotheses such as h*. In fact, Principle E could be employed to “confirm” all manner of nonscientific hypotheses more highly than competing scientific ones. Hypotheses postulating malignant fairies, vengeful ghosts, angry tribal gods, etc. can be shown on Principal E to be more likely than more mundane rivals.

Perhaps Schlesinger would reply by reminding us that he has specifically stated that even though the available evidence might better confirm an hypothesis than any of its rivals, the a priori likelihood of that hypothesis might be so low that one of the rivals is overall more probable.[44] Hence, hypotheses postulating demonds, fairies, witches, etc., though strictly speaking the evidence might confirm them more highly than scientific ones, are to be dismissed because of their extremely low initial credibility. However, Schlesinger gives no indication of how we are to arrive at such estimates of initial credibility. More significantly, given that hypotheses postulating demons, ghosts, hobgoblins, etc. are too a priori unlikely to be taken seriously, he does not say why the same should not apply to hypotheses invoking God.

Schlesinger’s Principle A tells us that when we are confronted with a multitude of hypotheses symmetrically related to the evidence, we are to employ the guiding rule that permits us to select a single, determinate hypothesis. With respect to scientific hypotheses, the correct guiding rule enjoins us to select the most parsimonious hypothesis. With respect to the choice between T and all the D-like hypotheses, the correct rule is to select the hypothesis postulating a perfect being. As we have seen, Schlesinger does not make it clear whether he has given us two different guiding rules–one for science and one for metaphysics–or whether the hypothesis postulating the perfect being is ipso facto the most parsimonious. With respect to this latter possibility, we have seen that the hypothesis of a perfect creator does not seem at all parsimonious given the nature of the evidence.

A deeper source of suspicion about Schlesinger’s guiding rules (assuming they are two distinct rules) is their algorithmic nature. Anyone versed in the postpositivist developments in the philosophy of science dealt with in the last chapter ought to suspect rules that promise immediate, automatic hypothesis selection. Surely it is a gross oversimplification of scientific procedure to prescribe the automatic selection of the most parsimonious available hypothesis. Schlesinger’s reason for insisting on this procedure, that the only way to get a single, determinate hypothesis is to select the most parsimonious one, is simply wrong. An hypothesis other than the most parsimonious one might be the one that best accords with previously accepted theory, or is richest in test implications, or promises solutions to the greatest number of other puzzles. In any such case parsimony would only be one and not necessarily the decisive selection principle. The specific hypothesis that becomes accepted will be the one that finally emerges from scientific debate as the one judged most plausible on the basis of a variety of criteria.

Difficulties also surround the application of Principle A to the selection of T from among all the D-like hypotheses. Schlesinger prescribes “select the perfect being” as a guiding rule because because he claims it is the only rule that picks out a necessarily unique being. This is not so, however. The following rule also appears to select a necessarily unique being: “Select the being who is maximally imperfect with respect to moral attributes and absolutely perfect with respect to all other predicates.” Unless there is some hidden incoherence in the conception of a being who is as perfect as the theistic god in power, knowledge, etc. but who is also completely evil, the above seems to be just as “adequate” a guiding rule as Schlesinger’s.

We have now seen that the arguments that Schlesinger bases on his Principles A and E are faulty. We have also seen that those Principles themselves are, at the very least, highly dubious and hence that no good arguments can be expected to derive from them. It therefore must be concluded that Schlesinger has failed in his attempt to show that the evidence makes theism more likely to be true than naturalism. He has also not shown that T is more likely to be true than any of the D-like hypotheses. We shall now turn to Schlesinger’s colleague, Richard Swinburne, to see if his effort fares any better.


[1] George Schlesinger, Religion and Scientific Method (Dordrecht, Holland: D. Reidel, 1977) and Metaphysics (Oxford: Basil Blackwell, 1983); Richard Swinburne, The Existence of God (Oxford: Clarendon Press, 1979).

[2] See Swinburne, pp. 277-291 and Schlesinger, Religion and Scientific Method, pp. 182-192.

[3] Metaphysics, p. 48.

[4] Ibid., pp. 13-16.

[5] Ibid., pp. 29-46.

[6] Ibid., pp. 21-22.

[7] Ibid., pp. 18-19.

[8] Schlesinger’s treatment of the miraculous is found in Religion and Scientific Method, pp. 173-181 and Metaphysics pp. 83-94.

[9] Religion and Scientific Method, p. 156.

[10] Ibid., p. 157.

[11] Ibid., pp. 157-158.

[12] Ibid., p. 158.

[13] Ibid., p. 161.

[14] Ibid., p. 163.

[15] Ibid., p. 163.

[16] P.F. Strawson, Introduction to Logical Theory, (London: Methuen and Co. Ltd., 1952), p. 257. Schlesinger endorses the Strawsonian approach in Religion and Scientific Method pp. 170-171.

[17] Religion and Scientific Method, p. 170.

[18] Ibid., p. 182.

[19] Ibid., p. 184.

[20] Metaphysics, pp. 73-74.

[21] Ibid., pp. 73-77.

[22] Ibid., p. 75.

[23] Ibid., pp. 77-78.

[24] Ibid., p. 78.

[25] Religion and Scientific Method, pp. 185-192.

[26] Ibid., pp. 191-192.

[27] Metaphysics, pp. 79-80.

[28] Flew, p. 98.

[29] See especially Kai Nielsen, Contemporary Critiques of Religion (London: Macmillan, 1971).

[30] Religion and Scientific Method, p. 157.

[31] Ibid., pp. 9-79 and Metaphysics pp. 48-66.

[32] Michael Martin, “Does the Evidence Confirm Theism More than Naturalism?” International Journal for Philosophy of Religion, 16 (1984), p. 258.

[33] Swinburne, p. 7.

[34] Religion and Scientific Method, p. 158.

[35] See for instance Mackie, p. 100.

[36] See Paul Davies, God and the New Physics (New York: Penguin Books, 1984), p. 171.

[37] Metaphysics, p. 72.

[38] See the definition of “Naturalism” in Antony Flew, ed. A Dictionary of Philosophy (New York: St. Martin’s, 1979).

[39] Ibid., p. 317.

[40] Flew gives a full defense of the Stratonician Presumption in God, Freedom, and Immortality (Buffalo: Prometheus, 1984), pp. 13-30.

[41] Metaphysics, p. 39.

[42] Ibid., p. 39.

[43] This example comes from Martin, p. 260.

[44] Religion and Scientific Method, pp. 158-159.