Naturalism and Supernaturalism as Frameworks

The previous rung argued that belief in God need not rest on evidence. This one takes the opposite approach seriously and asks how theism performs when it is treated as an explanatory hypothesis in direct competition with its rival.

That framing is not neutral, and both sides have reasons to dislike it. Classical theists object that God is not a hypothesis, because a hypothesis is a proposal about what one of the things in the world is like, and rung 5 recorded at length why they think that misdescribes the subject. Some naturalists object that theism is not a hypothesis either, because it predicts nothing and can accommodate anything, and an unfalsifiable claim is not a competitor. But the framing is where most contemporary argument actually happens, on both sides, and the remaining rungs of this tier are conducted almost entirely within it.

The two frameworks

The definitional difficulty on the naturalist side is worth dwelling on, because it is not merely pedantic. If naturalism means whatever science ends up recognising, it is unfalsifiable in the way naturalists accuse theism of being. If it means something more specific, such as the thesis that everything is composed of the entities of current physics, it is very likely false, since physics is unfinished. Most careful naturalists therefore state the position negatively, as the denial of a specific list, which is honest but concedes that the position is defined by what it rejects rather than by what it asserts.

The methodological version

A great deal of confusion in public argument about science and religion comes from running together two very different theses that share a word.

The distinction matters legally as well as philosophically. In the American litigation over teaching intelligent design, the defence of methodological naturalism turned on precisely this point: that excluding supernatural explanations from science does not amount to asserting atheism, and that a great many working scientists who accept the rule in the laboratory are religious believers outside it.

The philosophical case for the rule is that supernatural hypotheses are not tractable by scientific means. An agent who can do anything places no constraints on what will be observed, so no observation confirms or disconfirms the hypothesis, and a discipline built around controlled prediction has nothing to do with it. The rule is then not a prejudice but a recognition of what the method can handle.

The objection is that this converts a limitation into a boundary. If a supernatural cause were operating, methodological naturalism would guarantee that science never found it, and a method that cannot detect a certain kind of truth is not thereby entitled to conclude that no such truth exists. Critics add that the rule is often quietly upgraded: the claim that science does not consider supernatural causes slides into the claim that there are none, and the second does not follow from the first.

The standard reply concedes the logical point and defends the rule on pragmatic grounds. Nothing about the rule prevents philosophy, as distinct from science, from considering supernatural explanations, and this rung is an instance of philosophy doing so. What the rule secures is that the sciences do not stall: an inquiry permitted to answer "an agent did it" has a universally available answer and therefore a standing temptation to stop.

Explanation as a form of argument

The pattern is not a philosopher's invention. It is how the existence of atoms was established, how continental drift was accepted, and how anyone reasons from a wet umbrella to rain. And it is what nearly every contemporary argument for theism actually is, once the deductive packaging is removed. The teleological argument in its modern form does not claim that order entails a designer; it claims that a designer explains the order better than the alternatives. The fine-tuning argument does not claim that the constants could not have been what they are; it claims that their being life-permitting is much more expected on theism than on chance.

The pattern has a known weakness, and it applies symmetrically to both frameworks. Best is relative to the candidates considered, so an inference to the best explanation is only as good as the field of alternatives, and the history of science is full of cases where the best available explanation was the best of a bad set. Lipton's response is that we are reasonably good at generating candidates and that the objection, pressed hard, would undermine most of what we take ourselves to know. That defence works better in settings where the field of candidates is disciplined by experiment than in metaphysics, where it is disciplined only by imagination.

Setting out the comparison formally makes clear how much of the work is done before any evidence is examined.

Three features of that structure are worth noting before anything is put into it. The prior probabilities are not measurable, so the dispute over simplicity below is really a dispute about what to assign there. The expectations are not measurable either, and are frequently set by intuition dressed up as a judgement about what a perfect being would do. And the requirement that the ledger be total is honoured far more often in statement than in practice, since both sides tend to publish on the items favourable to them.

Richard Swinburne's The Existence of God (Clarendon Press, 1979; second edition 2004) is the most systematic attempt to run the theistic case this way. Two moves in it are worth isolating, because most of the argument since has been about them.

The first is personal explanation. Swinburne argues that we recognise two irreducible kinds of explanation: scientific, in terms of prior states and laws, and personal, in terms of an agent's intentions and powers. Neither reduces to the other, and appeals to agents are ordinary and respectable everywhere else. Theism is then a personal explanation of the most general features of the world, and the question is whether it is a good one, not whether that kind of explanation is permitted.

The second is the claim that theism is a simple hypothesis. This looks obviously wrong at first: a person with unlimited power, knowledge, and goodness sounds like the most extravagant posit imaginable. Swinburne's argument is that simplicity is a matter of the number and arbitrariness of the quantities a hypothesis specifies, and that infinite values are simpler than large finite ones, since a finite value raises the question of why that value rather than a slightly different one. A being of unlimited power fixes no arbitrary constants, and by that measure counts as simple.

That exchange is the state of play. J. L. Mackie's The Miracle of Theism (Clarendon Press, 1982) presses the objection in a sustained form, and Swinburne replies in the second edition. The dispute has not been resolved, and it is one of the places where the appearance of a technical disagreement conceals a difference in what the two sides think explanation is for.

Whether the explanation needs explaining

A recurring objection to any theistic explanation is that it does not explain anything, because the explanation is at least as puzzling as what it was called in to account for. If the order of the universe requires a cause, the mind that produced that order requires one too, and the manoeuvre has bought nothing.

The reply available to the theist is a general point about explanation, not a special plea. Explanations do not have to be explained in order to explain. Nobody objects that the kinetic theory of gases fails because the existence of molecules is unexplained. Every explanatory chain terminates somewhere, and the terminus is unexplained by definition, so demanding that the terminus be explained is demanding that there be no explanations at all.

The naturalist's version of the point is that this is exactly why the universe itself can be the terminus. If something has to be a brute fact, the most economical candidate is the thing we already know exists, rather than an additional entity posited behind it. Both frameworks end in something unexplained, and the disagreement is over which brute fact is easier to accept.

The theist's answer is that not every terminus is equally satisfactory, and that the difference lies in contingency. A brute fact that could have been otherwise leaves a question standing that a necessary terminus does not, and the terminus of rung 3 was necessary in a strong sense. The naturalist replies that necessity of that kind has never been shown to be coherent, and rung 3 recorded that dispute. What the framework framing adds is only that the disagreement is not local: it turns on what each side already thinks explanation is supposed to deliver.

What each framework has to absorb

Comparison is symmetric, and a comparison that lists only the costs of one side is not a comparison. Each framework has a set of features of the world that sit awkwardly within it, and where the standard move is to argue that the feature is less puzzling than it looks or will be accounted for eventually.

For theism, the awkward items are the ones the rest of this tier is about: the amount and distribution of suffering, the existence of thoughtful people who have looked and not found, and the availability of accounts of religious belief that do not require the belief to be true. Each of these is evidence the naturalist expects and the theist has to accommodate.

For naturalism, the standard list is consciousness, intentionality, mathematics, and moral facts. Conscious experience is hard to place in a purely physical account, and the difficulty is not merely that the account is unfinished but that it is unclear what a finished one would look like. Intentionality, the fact that a thought is about something beyond itself, has the same shape. Abstract mathematical truths appear to hold independently of the physical world while somehow applying to it. And if moral obligations are objective, they are a kind of fact that has no obvious home in the natural order, which is the ground the moral argument of the intermediate tier occupied.

The naturalist responses are not weak, and each has a substantial literature: consciousness may be a physical process we do not yet understand, mathematics may be a useful fiction rather than a description of anything, moral facts may be constructed rather than discovered. But they are accommodations, and they have costs. A theist who is told that theism explains nothing because it can accommodate anything is entitled to ask how these differ.

What the symmetry establishes is limited but real. Neither framework confronts a decisive counterexample, and both handle their awkward cases by argument rather than by capitulation. That is why the productive disputes are quantitative rather than qualitative, and why the machinery of likelihoods keeps appearing.

The gaps

The criticism has real force and a track record behind it. Planetary motion, the origin of species, the nature of disease, and the formation of the solar system were each at some point taken to require special divine action and are now accounted for without it. Induction over that record suggests that current gaps will close too, and that betting on a gap has been a losing strategy for four centuries.

The theistic response is that the criticism proves too much if it is applied indiscriminately, because it would rule out any inference to an agent from unexplained data, including the ones we accept. Finding an unexplained regularity in a signal from space and inferring a sender is not a fallacy; it is inference to the best explanation.

What the theist needs, and what the criticism demands, is a distinction between two kinds of gap. A contingent gap is a hole in current knowledge that further inquiry might fill, and inferring an agent from one of those is exactly the mistake Drummond described. A principled gap is one where the reason for the gap is a feature of the phenomenon rather than of our present ignorance: where we can say in advance why no explanation of the relevant kind could be forthcoming.

Whether any principled gaps exist is the live question. The fine-tuning argument is the strongest candidate, because its defenders argue that the values of the constants are not the sort of thing a deeper physics could explain without simply relocating the question to whatever fixed the deeper physics. Their opponents reply that this is precisely what was said about the orbits of the planets, and that the multiverse hypothesis shows how a deeper physics might dissolve the problem. Miracles are a weaker candidate for present purposes, since a single anomalous event is by its nature the sort of thing further investigation might explain, and rung 10 of the intermediate tier examined why the evidential situation there is unfavourable.

There is a theological version of the complaint that cuts deeper than the scientific one, and it is Bonhoeffer's. A God posited to fill explanatory gaps is a God whose relevance shrinks as knowledge grows, and the classical tradition of rung 4 never held such a view: on that account God is the reason there is a law-governed order at all, not an intervention within it. A theist committed to gap-filling has, on this view, already accepted the personalist picture and its liabilities, and is fighting on ground the tradition would not have chosen.

Adding things up

Almost nobody argues that a single consideration settles the question. The standard structure is a cumulative case argument, in which cosmological, teleological, moral, and experiential considerations each contribute something and the conclusion rests on the accumulation.

The natural objection is that adding weak arguments together yields nothing, and it is often put by saying that a number of leaky buckets hold no more water than one. If each argument fails, the conjunction of failures is not a success.

The Bayesian answer is that the image is wrong about how evidence works. Items of evidence that are independent of one another do accumulate: each shifts the probability by its own factor, and the shifts compound. Several considerations, none decisive, can jointly raise a probability substantially, and this is entirely ordinary in law and in medicine. Swinburne's treatment is explicitly of this form, and the intermediate tier introduced the machinery.

Two conditions have to be met for the reply to work, and both are contested.

The evidence has to be genuinely independent. If the considerations all trace to one underlying feature of the world, then they are one item of evidence described several ways, and counting them separately double-counts.

And the accumulation has to include the evidence on the other side. A cumulative case that gathers everything favourable and omits everything unfavourable is not a probability calculation but a brief. The remaining rungs of this tier are largely occupied with what belongs on the other side of the ledger: the distribution of suffering, the failure of many to believe, the availability of naturalistic accounts of religious belief. A serious cumulative case has to run in both directions, and this is where the framework framing earns its place, since a total assessment is exactly what comparing two frameworks requires.

What the framing settles and what it hides

Treating the two positions as rival explanations of shared data has a consequence that is easy to miss and that shapes every later argument in this tier.

Under it, no single item of evidence is decisive, and neither side has to explain anything in isolation. Each has a background picture that determines what counts as surprising. The theist who finds fine-tuning astonishing and the naturalist who finds it unremarkable are not disagreeing about the physics; they are disagreeing about what should have been expected, and expectation is set by the framework. That is why the arguments do not converge, and it is why the productive disputes in contemporary philosophy of religion are the ones that identify evidence whose expectedness differs sharply between the two.

That is exactly what the next rung supplies, which is why it comes next.

Next

The most careful evidential argument against theism does not claim that any suffering is pointless, which is a claim that invites the reply that we would not know. It claims something narrower and harder to answer: that the pattern in which pain and pleasure are distributed among living things tracks biological usefulness with a precision that theism gives no reason to expect and that indifference does. The next rung sets out that argument, its likelihood structure, and the theistic responses.