Skip to content

Scientific Naturalism Thesis Statement

1. Ontological Naturalism

1.1 Making a Causal Difference

A central thought in ontological naturalism is that all spatiotemporal entities must be identical to or metaphysically constituted by physical[3] entities. Many ontological naturalists thus adopt a physicalist attitude to mental, biological and other such “special” subject matters. They hold that there is nothing more to the mental, biological and social realms than arrangements of physical entities.

The driving motivation for this kind of ontological naturalism is the need to explain how special entities can have physical effects. Thus many contemporary thinkers adopt a physicalist view of the mental realm because they think that otherwise we will be unable to explain how mental processes can causally influence our bodies and other physical items. Similar considerations motivate ontologically naturalist views of the biological realm, the social realm, and so on.

It may not be immediately obvious why this need to account for physical effects should impose any substantial naturalist constraints on some category. After all, there seems nothing a priori incoherent in the idea of radically unscientific “supernatural” events exerting a causal influence on physical processes, as is testified by the conceptual cogency of traditional stories about the worldly interventions of immaterial deities and other outlandish beings.

However, there may be a posteriori objections to such non-natural causal influences on the physical world, even if there are no a priori objections. We shall see below how modern scientific theory places strong restrictions on the kinds of entities that can have physical effects. Given that mental, biological and social phenomena do have such effects, it follows that they must satisfy the relevant restrictions.

Note how this kind of argument bites directly only on those categories that do have physical effects. It places no immediate constraints on categories that lack any such effects, which arguably include the mathematical and modal realms, and perhaps the moral realm. We shall return to the question of whether there are any further reasons for ontologically naturalist views about causally non-efficacious categories in sections 1.7 and 1.8 below.

1.2 Modern Science and Causal Influence

There is an interesting history to modern science’s views about the kinds of things that can produce physical effects. It will be worth rehearsing this history in outline, if only to forestall a common reaction to ontological naturalism. It is sometimes suggested that ontological naturalism rests, not on reasoned argument, but on some kind of unargued commitment, some ultimate decision to nail one’s philosophical colours to the naturalist mast.[4] And this diagnosis seems to be supported by the historical contingency of naturalist doctrines, and in particular by the fact that they have become widely popular only in the past few decades. However, familiarity with the relevant scientific history casts the matter in a different light. It turns out that naturalist doctrines, far from varying with ephemeral fashion, are closely responsive to received scientific opinion about the range of causes that can have physical effects.

A short version of this history runs like this: (1) the mechanistic physics of the seventeenth century allowed only a very narrow range of such causes; (2) early Newtonian physics was more liberal, and indeed did not impose any real restrictions on possible causes of physical effects; (3) however, the discovery of the conservation of energy in the middle of the nineteenth century limited the range of possible causes once more; (4) moreover, twentieth-century physiological research has arguably provided evidence for yet further restrictions.

Let us now rehearse this story more slowly.

(1) The “mechanical philosophers” of the early seventeenth century held that any material body maintains a constant velocity unless acted on, and moreover held that all action is due to impact between one material particle and another. So stated, the mechanical philosophy immediately precludes anything except impacting material particles from producing physical effects. Leibniz saw this clearly, and concluded that it discredited Descartes’ interactive dualism, which had a non-material mind influencing the physical world (Woolhouse 1985). (Of course, Leibniz did not therewith reject dualism and embrace the physicalist view that minds are composed of material particles, but instead opted for “pre-established harmony”. Views which avoid ontological naturalistic views of the mind by denying that it has any physical effects will be discussed further in section 1.6 below.)

(2) At the end of the seventeenth century Newtonian physics replaced the mechanical philosophy of Descartes and Leibniz. This reinstated the possibility of interactive dualism, since it allowed disembodied forces as well as impacts to cause physical effects. Newtonian physics was open-ended about the kinds of forces that exist. Early Newtonians posited fundamental mental and vital forces alongside magnetic, chemical, cohesive, gravitational and impact forces. Accordingly, they took sui generis mental action in the material world to be perfectly consistent with the principle of physics. Moreover, there is nothing in the original principles of Newtonian mechanics to stop mental forces arising autonomously and spontaneously, in line with common assumptions about the operation of the mind (Papineau 2002: Appendix Section 3).

(3) In the middle of the nineteenth century the conservation of kinetic plus potential energy came to be accepted as a basic principle of physics (Elkana 1974). In itself this does not rule out fundamental mental or vital forces, for there is no reason why such forces should not be “conservative”, operating in such a way as to compensate losses of kinetic energy by gains in potential energy and vice versa. (The term “nervous energy” is a relic of the widespread late nineteenth-century assumption that mental processes store up a species of potential energy that is then released in action.) However, the conservation of energy does imply that any such special forces must be governed by strict deterministic laws: if mental or vital forces arose spontaneously, then there would be nothing to ensure that they never led to energy increases.

(4) During the course of the twentieth century received scientific opinion became even more restrictive about possible causes of physical effects, and came to reject sui generis mental or vital causes, even of a law-governed and predictable kind. Detailed physiological research, especially into nerve cells, gave no indication of any physical effects that cannot be explained in terms of basic physical forces that also occur outside living bodies. By the middle of the twentieth century, belief in sui generis mental or vital forces had become a minority view. This led to the widespread acceptance of the doctrine now known as the “causal closure” or the “causal completeness of the physical”, according to which all physical effects have fully physical causes.

1.3 The Rise of Physicalism

This historical sequence casts light on the evolution of ontologically naturalist doctrines. In the initial seventeenth-century mechanical phase, there was a tension, as Leibniz observed, between the dominant strict mechanism and interactive dualism. However, once mechanism was replaced by the more liberal doctrines of Newtonian physics in the second phase, science ceased to raise any objections to dualism and more generally to non-physical causes of physical effects. As a result, the default philosophical view was a non-naturalist interactive pluralism which recognized a wide range of fundamental non-physical influences, including spontaneous mental influences (or “determinations of the soul” as they would then have been called).

In the third phase, the nineteenth-century discovery of the conservation of energy continued to allow that sui generis non-physical forces can interact with the physical world, but required that they be governed by strict force laws. sui generis mental and vital forces were still widely accepted, but an extensive philosophical debate about the significance of the conservation of energy led to a widespread recognition that any such forces would need to be law-governed and thus amenable to scientific investigation. We might usefully view this as a species of ontological naturalism that falls short of full physicalism. Mental and other special forces were non-physical in the sense that they arose only in special circumstances and not throughout the spatiotemporal realm, but even so they fell within the realm of scientific law and lacked spontaneous autonomy. (As many commentators at the time recognized, this weaker form of naturalism already carried significant philosophical implications, particularly for the possibility of free will.[5])

In the final twentieth-century phase, the acceptance of the casual closure of the physical led to full-fledged physicalism. The causal closure thesis implied that, if mental and other special causes are to produce physical effects, they must themselves be physically constituted. It thus gave rise to the strong physicalist doctrine that anything that has physical effects must itself be physical.

In support of this understanding of the twentieth-century history, it is noteworthy how philosophers began to formulate arguments for physicalism from the 1950s onwards. Some of these arguments appealed explicitly to the causal closure of the physical realm (Feigl 1958; Oppenheim and Putnam 1958). In other cases, the reliance on causal closure lay below the surface. However, it is not hard to see that even in these latter cases the causal closure thesis played a crucial role.

Thus, for example, consider J.J.C. Smart’s (1959) thought that we should identify mental states with brain states, for otherwise those mental states would be “nomological danglers” which play no role in the explanation of behaviour. Or take David Lewis’s (1966) and David Armstrong’s (1968) arguments that, since mental states are picked out by their causal roles, and since we know that physical states play these roles, mental states must be identical with those physical states. Again, consider Donald Davidson’s (1970) argument that, since the only laws governing behaviour are those connecting behaviour with physical antecedents, mental events can only be causes of behaviour if they are identical with those physical antecedents. At first sight, it might not be obvious that these arguments require the causal closure thesis. But a moment’s thought will show that none of these arguments would remain cogent if the closure thesis were not assumed, and it were thus left open that some physical effects (the movement of matter in arms, perhaps, or the firings of the motor neurones which instigate those movements) were not determined by prior physical causes at all, but by sui generis mental causes.

Sometimes it is suggested that the indeterminism of modern quantum mechanics creates room for sui generis non-physical causes to influence the physical world. However, even if quantum mechanics implies that some physical effects are themselves undetermined, it provides no reason to doubt a quantum version of the causal closure thesis, to the effect that the chances of those effects are fully fixed by prior physical circumstances. And this alone is enough to rule out sui generis non-physical causes. For such sui generis causes, if they are to be genuinely efficacious, must presumably make an independent difference to the chances of physical effects, and this in itself would be inconsistent with the quantum causal closure claim that such chances are already fixed by prior physical circumstances. Once more, it seems that anything that makes a difference to the physical realm must itself be physical.

1.4 Reductive and Non-Reductive Physicalism

It will be worth being explicit about the way the causal closure principle supports physicalism. First we assume that mental causes (biological, social, …) have physical effects. Then the causal closure principle tells us that those physical effects have physical causes. So, in order to avoid an unacceptable proliferation of causes for those physical effects (no “systematic overdetermination”), we need to conclude that the mental (biological, social, …) causes of those effects are not ontologically separate from their physical causes.

However, even if this general line of argument is accepted, there is room for differing views about exactly what its denial of ontological separateness requires. Let us agree that causes are “events” (or “facts”) that involve instantiations of properties.[6] So, if some special cause is not ontologically separate from some physical cause, the property instantiations that it involves cannot themselves be ontologically separate from the property instantiations involved in the physical cause. At this point, however, there are divergent views about how tight a constraint this imposes.

One school holds that it requires type-identity, the strict identity of the relevant special properties with physical properties. On the other side stand “non-reductive” physicalists, who hold that the causal efficacy of special causes will be respected as long as the properties they involve are “realized by” physical properties, even if they are not reductively identified with them.

Type-identity is the most obvious way to ensure the non-separateness of special and physical causes: if exactly the same properties comprise the special and physical cause, the two causes will themselves be fully identical. Still, type-identity is a very strong doctrine. Type identity about thoughts, for example, would imply that the property of thinking about the square root of two is identical with some physical property. And this seems highly implausible. Even if all human beings with this thought must be distinguished by some common physical property of their brains—which itself seems unlikely—there remains the argument that other life-forms, or intelligent androids, will also be able to think about the square root of two, even though their brains may share no significant physical properties with ours (Fodor 1974; Bickle 2013).

This “variable realization” argument has led many philosophers to seek an alternative way of reconciling the efficacy of mental and other special causes with the causal closure thesis, one which does not require the strict identity of non-physical and physical properties. The general idea of this “non-reductive physicalism” is to allow that instantiations of a given special property will always be grounded in or metaphysically determined by instantiations of physical properties, but that these “realizing” physical properties might be different in different cases. So, for example, any being who thinks about the square root of two will do so in virtue of instantiating some physical properties, but these can be different physical properties in different cases—in one human being it may be one set of neural arrangements, in another a different set, and in other life forms it might involves nothing like neural properties at all.

There are various more detailed ways of filling out this idea of non-reductive physicalism. A common feature is the requirement that special properties should metaphysically supervene on physical properties, in the sense that any two beings who share the realizing physical properties will necessarily share the same special properties, even though the physical properties which so realize the special ones can be different in different beings. This arguably ensures that nothing more is required for any specific instantiation of a special property than its physical realization—even God could not have created your brain states without thereby creating your feelings—yet avoids any reductive identification of special properties with physical ones. (This is a rough sketch of the supervenience formulation of physicalism. For more see Stoljar 2015.)

Some philosophers object that non-reductive physicalism does not in fact satisfy the original motivation for physicalism, on the grounds that it does not really reconcile the efficacy of mental and other special causes with the causal closure thesis (Kim 1998; Robb and Heil 2014: Section 6). According to non-reductive physicalism, special properties are not type-identical with any strictly physical properties, even though they supervene on them. But this then seems to imply that any given special cause will be distinguishable from the physical cause that realizes it, to the extent that it involves the instantiation of a different property. (The property of thinking about the square root of two is definitely a different property from the neural property that realizes it in me, say, since another being could share the former property without sharing the latter.)

The opponents of non-reductive physicalism then insist that this gives us an unacceptable proliferation of causes for the physical effects of special causes after all—both the physical cause implied by the causal closure thesis and the distinct special cause. In response, advocates of non-reductive physicalism respond that there is nothing wrong with such an apparent duplication of causes if it is also specified that the latter metaphysically supervene on the former.

The issue here hinges on the acceptability of different kinds of systematic overdetermination (Bennett 2003). All can agree that it would be absurd if the physical effects of special causes always had two metaphysically independent causes. Plugging this into the causal closure argument for physicalism, we can conclude that there can be no metaphysically independent non-physical causes (such as Cartesian dualist mental causes) for effects that already have full physical causes. However, even if “strong overdetermination” by two ontologically independent causes is so ruled out, this does not necessarily preclude “weak overdetermination” by both a physical cause and a metaphysically supervenient special cause. Advocates of non-reductive physicalism argue that this kind of overdetermination is benign and consistent with the causal argument’s denial of strong overdetermination, since now the two causes are not metaphysically distinct—the special cause isn’t genuinely additional to the physical cause (nothing more is needed for your feelings than your brain states).

1.5 Physicalist Downwards Causation

Some recent writers have explored a different way of upholding the causal efficacy of non-reduced mental and other special causes. Where the “benign overdetermination” option says certain effects have a special cause as well as a physical cause, these writers urge that some effects have a special cause instead of a physical cause.

Suppose a pigeon pecks at crimson tiles. Is the pecking caused by the specific shade, crimson, or the more generic colour, red? The natural answer is that it depends. If the pigeon pecks only at crimson tiles, and not at other shades of red, then it is the crimsonness that is causing the pecking, whereas, if the pigeon pecks at any shade of red, it is the redness. Examples like these have led a number of writers to require that causes be proportional to their effects (Yablo 1992; Menzies 2008; List and Menzies 2009, 2010). We should attribute the effect to that property that is specific enough to suffice for it, but no more specific than that.

This suggests that sometimes special causes and not their physical realizers might be responsible for physical effects. Suppose I want to hail a taxi, and that this desire is realized by some brain state, and that I then wave my arm. It seems that it will then be the desire that is proportional to the waving, not the brain state, in that I would still have waved my arm if my desire had been differently realized, though not if I hadn’t had the desire at all.

Some will say that in such cases the desire causally explains the waving, but that it is still the brain state that causes it. This thought appeals to the intuition that real causal relations are always constituted by basic physical interactions, by bits of matter bumping into each other. But this intuition is not decisive, and a number of theoretical considerations speak against it. For instance, the “difference-making” account of causation developed by James Woodward implies that generic properties can often eclipse their more specific realizations as causes (Woodward 2005), as does the view that causation is a relatively macroscopic phenomenon whose temporal asymmetry is analogous to the temporal asymmetries of thermodynamics (Loewer 2007; Papineau 2013).

It is worth observing that physicalists who advocate this kind of downwards special causation are in some danger of sawing off the branch they are sitting on, in that they now seem to be advocating counter-examples to the causal closure of the physical. If my arm’s waving is caused by my desire and not by my brain state, it has a special cause and not a physical cause, and this will refute the closure thesis that every physical effect has some fully physical cause. Since the original rationale for embracing physicalism was supposed to be science’s discovery that the physical realm is causally closed, this may seem to leave physicalists in an awkward position.

However, even if the examples of “downwards” causation do undermine the thesis of the causal closure of the physical, it may still be possible to rework the original argument for physicalism in terms of closure under nomological determination, rather than causal closure.[7] Nothing in the idea of proportional causation threatens the idea that modern physics shows that all physical effects (or their chances) are fully nomologically determined by physical antecedents, in the form of non-special force fields, even if they aren’t always caused by them. And it is arguable that this in itself rules out the possibility of some metaphysically independent non-physical factor making any causal difference to the unfolding of the physical world.

1.6 Conscious Properties and the Causal Closure Arguments

Some philosophers take there to be compelling arguments against the view that conscious states are metaphysically constituted by physical states. Putting this together with the closure claim that physical effects always have physical causes, and abjuring the idea that the physical effects of conscious causes might be strongly overdetermined by both a physical cause and an ontologically independent conscious cause, they are led to the view that conscious states must be “epiphenomenal”, lacking any power to causally influence the physical realm (Jackson 1982, 1986; Chalmers 1996).[8]

The rejection of physicalism about conscious properties certainly has the backing of intuition. (Don’t zombies—beings who are physically exactly like humans but have no conscious life—intuitively seem metaphysically possible?) However, whether this intuition can be parlayed into a sound argument is a highly controversial issue, and one that lies beyond the scope of this entry. A majority of contemporary philosophers probably hold that physicalism can resist these arguments.[9] But a significant minority take the other side, and for them epiphenomenalism is the natural option.[10]

If the majority are right, and the arguments against physicalism about conscious states are not in fact compelling, then physicalism seems clearly preferable to epiphenomenalism. In itself, epiphenomenalism is not an attractive position. It requires us to suppose that conscious states, even though they are caused by processes in the physical world, have no effects on that world. This is a very odd kind of causal structure. Nature displays no other examples of such one-way causal intercourse between realms. By contrast, a physicalist naturalism about conscious states will integrate the mental realm with the causal unfolding of the spatiotemporal world in an entirely familiar way. Given this, general principles of theory choice would seem to argue strongly for physicalism over epiphenomenalism.[11]

If we focus on this last point, we might start wondering why the causal closure thesis is so important. If general principles of theory choice can justify physicalism, why bring in all the complications associated with causal closure? The answer is that causal closure is needed to rule out interactionist dualism. General principles of theory choice may dismiss epiphenomenalism in favour of physicalism, but they do not similarly discredit interactionist dualism. As the brief historical sketch earlier will have made clear, interactionist dualism offers a perfectly straightforward theoretical option requiring no commitment to any bizarre causal structures. Certainly the historical norm has been to regard it as the default account of the causal role of the mental realm. Given this, arguments from theoretical simplicity cut no ice against interactionist dualism. Rather, the case against interactionist dualism hinges crucially on the empirical thesis that all physical effects already have physical causes. It is specifically this claim that makes it difficult to see how dualist states can make a causal difference to the physical world.[12]

It is sometimes suggested that physicalism about the mind can be vindicated by an “inference to the best explanation”. The thought here is that there are many well-established synchronic correlations between mental states and brain states, and that physicalism is a “better explanation” of these correlations than epiphenomenalism (Hill 1991; Hill and McLaughlin 1999). From the perspective outlined here, this starts the argument in the middle rather than the beginning, in effect assuming that interactionist dualism has already been ruled out. After all, if we believed interactionist dualism, then we wouldn’t think that fundamentally dualist mental pains, say, needed any help from synchronic neural correlates to produce physical effects, and so wouldn’t have any reason to suppose that they were synchronically correlated with any particular brain states. Rather our rationale for believing in such synchronic correlations is that the causal closure of the physical realm has already ruled out interactive dualism, so to speak, whence we infer that mental states can only systematically precede physical effects if they are correlated with the physical causes of those effects.

1.7 Moral Facts

G.E. Moore’s well-known “open question” argument is designed to show that moral facts cannot possibly be identical to natural facts. Suppose that the natural properties of some situation are completely specified. It will always remain an open question, argued Moore, whether that situation is morally good or bad (Moore 1903).

Moore took this argument to show that moral facts constitute a distinct species of non-natural fact. However, any such non-naturalist view of morality faces immediate difficulties, deriving ultimately from the kind of causal closure thesis discussed above. If all physical effects are due to a limited range of physically-grounded natural causes, and if moral facts lie outside this range, then it follow that moral facts can never make any difference to what happens in the physical world (Harman 1986). At first sight this may seem tolerable (perhaps moral facts indeed don’t have any physical effects). But it has awkward epistemological consequences. For beings like us, knowledge of the spatiotemporal world is mediated by physical processes involving our sense organs and cognitive systems. If moral facts cannot influence the physical world, then it is hard to see how we can have any knowledge of them.

The traditional non-naturalist answer to this problem is to posit a non-natural faculty of “moral intuition” that gives us some kind of direct access to the moral realm (as explained in Ridge 2014: Section 3). However, causal closure once more makes it difficult to make good sense of this suggestion. Presumably at some point the posited intuitive faculty will need to make a causal difference in the physical world (by affecting what people say and do, for example). And at this point the causal closure argument will bite once more, to show that a non-natural intuitive faculty would inevitably commit us to the absurdity of strong overdetermination.[13]

In the face of this kind of difficulty, most contemporary moral philosophers have turned away from moral non-naturalism and opted for some species of naturalist view. We can divide the naturalist options here into two broad categories: irrealist and realist. Irrealist moral naturalists aim to account for moral discourse by offering naturalist accounts of the social and linguistic and practices that govern it, but without supposing that moral utterances report on moral facts with a substantial independent existence (Joyce 2015). By contrast, naturalist moral realists agree with moral non-naturalists against irrealists that substantial moral facts exist, but seek to locate them in the natural realm rather than in some sui generis non-natural realm (Lenman 2014).

Both these broad categories have further sub-divisions. Among the irrealists, we can distinguish explicitly non-cognitivist views like emotivism and prescriptivism which deny that moral judgements express beliefs (Hare 1952; Blackburn 1993; Gibbard 2003), from cognitivist views that accept that moral judgements do express beliefs but deny a substantial reality to the putative facts to which they answer; and among the latter cognitivist views we can distinguish error-theoretic fictionalist options which view moral judgements as simply false (Mackie 1977; Kalderon 2005) from projectivist options which hold that moral discourse is sufficiently disciplined for its judgements to qualify for a species of truth even though they do not report on independently existing causally significant facts (Wright 1992; Price 2011).

Naturalist moral realism also comes in different varieties. In recent debates two versions have figured prominently; “Cornell realism”, which includes moral facts among the causally significant facts but resists their type-reducibility to non-moral facts (Sturgeon 1985; Boyd 1988), and “moral functionalism” which is happy to equate moral facts with straightforwardly descriptive facts (Jackson 1998).

Any kind of moral naturalist realist needs to reject Moore’s open question argument. They have two alternatives here. One is to insist that Moore’s posited openness is relatively superficial, and that there is no principled barrier to inferring moral facts a priori from the non-moral natural facts, even if such inferences will sometimes require a great deal of information and reflection. The other is to argue that the constitution of moral facts by non-moral natural facts is an a posteriori matter, akin to the relation between water and H2O, and that therefore Moore’s openness only points to a conceptual gap, not a metaphysical one (Ridge 2014: Section 2).

This sub-section has focused on morality. But there are other areas which arguably involve matters of value, such as theoretical and practical reason, aesthetics, and so on. The constraints placed on theories of moral facts by naturalist considerations will apply, mutatis mutandis, in these areas too, militating against theories that posit non-natural value-bearing facts and in favour of naturalist alternatives, of either a realist or irrealist stripe.

1.8 Mathematical Facts

Mathematics raises many of the same issues for ontological naturalism as morality. Mathematical claims typically involve a commitment to abstract objects like numbers and sets, eternal entities outside space and time. This might seem cogent at first sight, but once more epistemological difficulties quickly arise. Abstract objects can have no effects in the spatiotemporal world. How then can spatiotemporal being like ourselves come to know about them?

However the mathematical case does not fully parallel the moral one. One of the options in the moral case was naturalist realism, which reads moral claims as about natural facts which play causal roles in the spatiotemporal world. However, given the explicit commitment of mathematical claims to abstract objects without spatiotemporal location, this option does not seem available in the mathematical case (but see Maddy 1990). So we seem required to choose between naturalist irrealism or an ontologically non-naturalist realism that upholds the reality of non-spatiotemporal mathematical entities.

As in the moral case, one response to the epistemological challenges facing such a non-naturalist realism is to posit some faculty of intuition which gives us access to the abstract mathematical realm. Kurt Gödel arguably favoured a view along these lines (Parsons 1995). However, once more this only seems to push the problem back. There seems no good way for the posited faculty to bridge the causal gap between the abstract and the spatiotemporal realms.

An alternative version of non-naturalist realism aims to vindicate mathematical and modal claims as essential parts of our best overall theories of the world. According to this line of thought, defended by Hilary Putnam, our empirically best-supported scientific theories commit us to mathematical entities; ergo, we are entitled to believe in such entities (Putnam 1971).[14]

However, it is contentious whether our best-supported empirical theories do commit us to abstract mathematical entities. The most prominent version of naturalist irrealism about mathematics, Hartry Field’s fictionalism, disputes precisely this claim. According to Field, there are ways of constructing “nominalist” versions of scientific theories that avoid commitment to abstract mathematical objects. From Field’s point of view, these nominalized versions of our scientific theories are explanatorily superior to the “platonist” alternatives. Field argues that we do not have to regard mathematics itself as literally true in order to understand its use in science and other applications. Rather it can be viewed as a “useful fiction” which facilitates inferences between nominalistic scientific claims, but is not itself implicated in our most serious beliefs about the world (Field 1980, 1989).

Not all philosophers of mathematics are convinced that Fieldian nominalizations are available to replace all scientific references to abstract mathematical objects. In particular, some have argued that certain explanations of nominalist facts make essential reference to abstract objects (Baker 2005; Batterman 2010). In response, others have sought to show that there are in fact good nominalist explanations of the facts in question (Daly and Langford 2009; Butterfield 2011; Menon and Callender 2013). In any case, it is not clear that Field’s metaphysical stance requires a full execution of his nominalizing programme, as opposed to a case for its cogency: difficulties in constructing nominalist theories can always be attributed to limitations of human ingenuity rather than the reality of abstract mathematical objects (Leng 2013).

Perhaps the most popular contemporary alternative to fictionalism is the version of non-naturalist realism offered by the neo-Fregean thesis that abstract mathematical beliefs can be justified as analytic truths that follow from logic and certain meaning stipulations. The idea has been most fully developed in connection with arithmetic, where Crispin Wright has shown how Peano’s postulates can be derived within the framework of second-order logic from nothing except the Humean principle that the same number attaches to equinumerous concepts. According to Wright, this principle can be viewed as an implicit definition of our concept of number. If this is right, then it has indeed been shown that arithmetic, and therewith the existence of numbers as abstract objects, follows from logic and definition alone (Wright 1983; Hale and Wright 2003). There has also been some attempt to extend the programme to mathematical analysis (Shapiro 2000; Wright 2000).

One query that might be put to this programme is whether the Humean principle and analogous assumptions can really be viewed as analytic definitions. If they commit us to numbers and other abstract objects whose existence cannot established without them, then they are arguably doing more than definitions should. A related issue is whether the overall neo-Fregean position is properly viewed as realist. From its perspective, the role of abstract mathematical objects in the overall scheme of things seems to be exhausted by their making our mathematical statements true; given this, it might seem better to classify it as a species of irrealism (MacBride 2003).

To complete this discussion of ontological naturalism, let us briefly consider the realm of modality, understood as the subject matter of claims that answer to something more than actuality. Modality raises many of the same issues as mathematics, but the topic is complicated by the prior question of the content of modal claims, and in particular about whether they quantify over non-actual possible worlds. Whereas there is little dispute about the initial semantic analysis of mathematical claims—they purport to refer to abstract numbers, sets, functions and so on—there is somewhat less unanimity about the possible worlds analysis of modal claims (Nolan 2011b).

To the extent that modal claims do quantify over possible worlds, the ontological points made about mathematics apply will here too. Since non-actual worlds do not inhabit our spatiotemporal realm, an ontologically naturalist realism seems to be ruled out from the start. The remaining alternatives are irrealism or non-naturalist realism. The former alternative has been explored in recent years by modal fictionalists (Rosen 1990; Nolan 2011a). The options under the latter heading meet the same epistemological challenges as in the mathematical case: brute intuition faces causal problems; it is contentious whether we should take our best scientific theories to commit us to possible worlds; and, if modal knowledge is to be analytically a priori, on the model of mathematical neo-Fregeanism, then it is not obvious that it can take us to knowledge of possible worlds construed realistically.

2. Methodological Naturalism

2.1 Philosophy and Science

In what follows, “methodological naturalism” will be understood as a view about philosophical practice. Methodological naturalists see philosophy and science as engaged in essentially the same enterprise, pursuing similar ends and using similar methods.

In some philosophy of religion circles, “methodological naturalism” is understood differently, as a thesis about natural scientific method itself, not about philosophical method. In this sense, “methodological naturalism” asserts that religious commitments have no relevance within science: natural science itself requires no specific attitude to religion, and can be practised just as well by adherents of religious faiths as by atheists or agnostics (Draper 2005). This thesis is of interest to philosophers of religion because many of them want to deny that methodological naturalism in this sense entails “philosophical naturalism”, understood as atheism or agnosticism. You can practice natural science in just the same way as non-believers, so this line of thought goes, yet remain a believer when it comes to religious questions.

Not all defenders of religious belief endorse this kind of “methodological naturalism”. Some think that religious doctrines do make a difference to scientific practice, yet are defensible for all that (Plantinga 1996). In any case, this kind of “methodological naturalism” will not be discussed further here. Our focus will be on the relation between philosophy and science, not between religion and science.

It is uncontentious that philosophers differ widely in their initial attitudes to natural science. Some philosophers celebrate science, and seek out ways in which philosophy can be illuminated by it. Other philosophers view science with suspicion, and feel that any dependence on science somehow infringes the autonomy of philosophy. However, such temperamental differences need not themselves amount to differing views about the nature of philosophy. After all, even those philosophers who are suspicious of science will need to allow scientific findings are sometimes of philosophical significance—we need only think of the role that the causal closure of physics was shown earlier to play in the contemporary mind-body debate. And, on the other side, even the philosophical friends of science must admit that there are some differences at least between philosophy and natural science—for example, that philosophers do not generally gather empirical data in the way that scientists do.

If we want to isolate a serious debate about philosophical method, we will need to go beyond initial reactions to science and look at more specific methodological commitments. In order to focus what follows, let us thus understand methodological naturalism as asserting that at bottom philosophy and science are both concerned to establish synthetic knowledge about the natural world, and moreover to achieve this by a posteriori investigation.

Methodological naturalists in this sense will of course allow that there are some differences between philosophy and science. But they will say that these are relatively superficial. In particular, they will argue that they are not differences in aims or methods, but simply a matter of philosophy and science focusing on different questions. For one thing, philosophical questions are often distinguished by their great generality. Where scientists think about viruses, electrons or stars, philosophers think about spatiotemporal continuants, properties, causation and time. Categories like these structure all our thinking about the natural world. A corollary is that alternative theories at this level are unlikely ever to be decided between by some simple experiment, which is no doubt one reason that philosophers do not normally seek out new empirical data. Even so, the methodological naturalist will insist, such theories are still synthetic theories about the natural world, answerable in the last instance to the tribunal of a posteriori empirical data.

Not all philosophical questions are of great generality. Think of topics like weakness of will, the importance of originality in art, or the semantics of fiction. What seems to identify these as philosophical issues is that our thinking is in some kind of theoretical tangle, supporting different lines of thought that lead to conflicting conclusions. Progress requires an unravelling of premises, including perhaps an unearthing of implicit assumptions that we didn’t realize we had, and a search for alternative positions that don’t generate further contradictions. Here too a posteriori data are clearly not going to be crucial in deciding theoretical questions—often we have all the data we could want, but can’t find a good way of accommodating them. Still, methodological naturalists will urge, this once more doesn’t mean that a posteriori synthetic theories are not the aim of philosophy. An a posteriori synthetic theory unraveled from a tangle is still an a posteriori synthetic theory, even if no new observations or experimental findings went into its construction.

2.2 Philosophical Intuitions: Analytic or Synthetic?

An obvious objection to methodological naturalism appeals to the role that intuitions play in philosophical debate. Where scientists test their theories against the findings of observation and experiment, philosophers typically test theirs against intuitions. On the face of things, this argues that philosophy uses a distinctively a priori method of investigation.

However, this objection is not decisive. One issue is whether intuitions do play a central role in philosophical method. We shall discuss this further in section 2.5 below. But even if it is assumed that intuitions are central to philosophy, this is not necessarily inconsistent with methodological naturalism. For it is not to be taken for granted that philosophical intuitions have a distinctive a priori status. There is also the possibility that they answer in the end to a posteriori evidence, like the kind of intuitions at play in scientific thought experiments.

Before we turn to these issues, a prior question is whether philosophical intuitions should be viewed as analytic or synthetic. While there is an extensive contemporary debate about the importance of intuitions in philosophy (DePaul and Ramsey 1998; Knobe and Nicholls 2008; Pust 2014; Cappelen 2012), this question is surprisingly little asked.[15] No doubt this is a result of Quine’s doubts about the analytic-synthetic distinction. This is not the place to assess Quine’s arguments. But, whatever exactly they establish, they do not rule out all notions of analyticity that might allow us to press our question. For example, the notion of a statement whose truth follows from logic and definitions will serve well enough for our purposes (Papineau 2015; see also Russell 2008).[16] We shall understand analyticity in this way in what follows.

The choice between analyticity and syntheticity poses a familiar dilemma for friends of a priori intuitions. If philosophical intuitions are analytic, then this will plausibly render them a priori, but it will call their philosophical significance in question. By contrast, if they are synthetic, then they may well be of substantial philosophical significance, but then it will be less easy to understand how they can be a priori.

2.3 The Canberra Plan

A significant number of contemporary philosophers view philosophy as resting on “conceptual analysis”. At first sight, this might seem to commit them to the view that philosophy trades in analytic claims. However, as we shall see in the next section, it is by no means clear that all its advocates understand conceptual analysis in this way.

One influential school, however, is relatively explicit about the importance of analytic claims to philosophy. Led by David Lewis and Frank Jackson, and widely known as the “Canberra Plan”, this approach argues that philosophy starts with an initial analysis of concepts employed by everyday thought, such as free will, say, or knowledge, or moral value, or conscious experience (Jackson 1998). Once this has been done, philosophy can then turn to “serious metaphysics” to demonstrate how a limited number of ingredients (for example, physical ingredients) might satisfy these everyday concepts. This second stage is likely to appeal to synthetic a posteriori scientific knowledge about the fundamental nature of reality. But the purely analytic first stage, argues Jackson, also plays an essential part in setting the agenda for the subsequent metaphysical investigation.

An initial question about this programme relates to its scope. It is by no means clear that all, or indeed any, philosophically interesting concepts can be subject to the relevant kind of analysis. Jackson himself assumes that pretty much all everyday concepts can be analysed as equivalent to “the kind which satisfies such-and-such folk assumptions”.[17] But it is arguable that many everyday concepts are not so constituted, but rather have their semantic contents fixed by observational, causal or historical relations to their referents.

Still, we can let this point pass. Even if we suppose, for the sake of the argument, that a range of philosophically interesting everyday concepts do have their contents fixed in the way the Canberra programme supposes, there are further objections to its understanding of philosophical method.

Let us look a bit more closely at the posited initial agenda-setting stage of the Canberra programme. Upon closer examination, it is not clear that this makes any essential appeal to analytic knowledge. Defenders of the Canberra plan characteristically explain their strategy in terms of “Ramsey sentences”(e.g., Jackson 1998: 140). Suppose \(T(F)\) is the set of relevant everyday assumptions involving some philosophically interesting concept. For example, F may be the concept belief, and the assumptions in T may include “characteristically caused by perceptions”, “combines with desires to generate actions”, and “has causally significant internal structure”. Then the Ramsey sentence corresponding to \(T(F)\) is “\(\exists!\Phi (T(\Phi))\)”.[18] For the concept belief, this would say

There is some unique kind that is characteristically caused by perceptions, combines with desires to generate actions, and has causally significant internal structure.

The Canberra suggestion is then that, once we have articulated the relevant Ramsey sentence, we will be in a position to turn to serious metaphysics to identify the underlying nature of the F that plays the relevant role.

However, if this is the Canberra procedure, then there is no reason to think of it as appealing to analytic knowledge at any stage. A Ramsey sentence of the kind at issue says that there actually exists some entity satisfying certain requirements (there is a kind of state that is caused by perceptions …). Sentences like this make eminently synthetic and falsifiable claims. It is not a matter of definition that humans actually have internal states that play the causal role associated with the concept of belief. The Canberra strategy thus seems no different from the prescription that philosophy should start with the synthetic theories endorsed by everyday thought, and then look to our more fundamental theories of reality to see what, if anything, makes these everyday theories true. This seems entirely in accord with methodological naturalism—philosophy is in the business of assessing and developing synthetic theories of the world.

The crucial point here is that Ramsey sentences don’t define concepts like belief, but eliminate them. They give us a way of saying what our everyday theories say without using the relevant concept (there is a kind of state that is caused by perceptions …) If we want definitions, then we need “Carnap sentences”, not Ramsey sentences (Lewis 1970). The Carnap sentence corresponding to “\(\exists!\Phi(T(\Phi))\)” is “If \(\exists!\Phi(T(\Phi)\), then \(T(F)\)”. (If there is a kind of state that is caused by perceptions …, then it’s belief.) Carnap sentences can plausibly be viewed as akin to stipulations that fix the reference of the relevant concepts, and to that extent as analytic claims that can be known a priori. But this certainly does not mean that Ramsey sentences, which make substantial claims about the actual world, are also knowable via a priori analysis. (Note how you can accept a conditional Carnap sentence even if you reject the corresponding unconditional Ramsey sentence. You can grasp the folk concept of belief even if you reject the substantial folk theory of belief.)

Can’t defenders of the Canberra programme argue that it is the analytic Carnap sentences that are crucial in setting philosophical agendas, not the synthetic Ramsey sentences? But this seems wrong. We will want to know about the fundamental nature of belief if we suppose that there is a kind of state that is characteristically caused by perceptions, and so on. That is certainly a good motivation for figuring out whether and how the fundamental components of reality might constitute this state. But the mere fact that everyday thought contains a concept of such a state in itself provides no motivation for further investigation. (In effect, the function of a Carnap sentence is to provide a shorthand for talking about the putative state posited by the corresponding Ramsey sentence. It is hard to see how any important philosophical issues could hang on the availability of such a shorthand.)

To emphasize the point, consider the everyday concept of a soul, understood something that is present in conscious beings and survives death. This concept of a soul can be captured by the analytic Carnap sentence: “If certain entities inhabit conscious beings and survive death, then they are souls”. Accordingly, this Carnap sentence will be agreed by everybody who has the concept of soul, whether or not they believe in souls. Yet this Carnap sentence will not per se raise any interesting metaphysical questions for those who deny the existence of souls. These deniers won’t start wondering how the fundamental constituents of reality realize souls—after all, they don’t believe in souls. It is only those who accept the corresponding Ramsey sentence (“There are parts of conscious beings that survive death”) who will see a metaphysical issue here. Moreover, the Ramsey sentence will pose this metaphysical issue whether or not it is accompanied by some analytic Carnap sentence to provide some shorthand alternative terminology. In short, the methodological naturalist can insist that anybody interested in “serious metaphysics” should start by articulating the substantial existential commitments of our folk theories, as articulated in their synthetic Ramsey sentences. Any further analytic conceptual commitments add nothing of philosophical significance.

The point generalizes beyond the contrast between Ramsey and Carnap sentences. On reflection, it is hard to see why any purely definitional analytic truths should matter to philosophy. Synthetic everyday truisms can certainly be philosophically significant, and so their articulation and evaluation can play an important philosophical role. But there is no obvious motive for philosophy to concern itself with definitions that carry no implications about the contents of reality.

2.4 A Priori synthetic Intuitions?

Advocates of the Canberra plan are explicit that “conceptual analysis” is concerned with analytic truths. However, many other philosophers who advertise themselves as engaging in “conceptual analysis” do not take themselves to be committed to this implication.

Sometimes it is explicitly specified that “conceptual analysis” involves articulating synthetic theories and assessing them against a posteriori evidence. (A particularly clear version of this picture of conceptual analysis is given in Brandom 2001. See also Goldman 2007.) So understood, the view that philosophy is engaged in “conceptual analysis” seems perfectly consistent with methodological naturalism.

Other philosophers are also explicit that “conceptual analysis” issues in synthetic claims, but simultaneously regard it as a source of a priori knowledge (e.g., Jenkins 2008, 2012). This combination of views is less straightforward. In particular, it seems open to the traditional query: how is such synthetic a priori knowledge possible? If some claim is not guaranteed by the structure of our concepts, but answers to the nature of the world, then how is it possible to know it without a posteriori evidence?

Yet other philosophers distance themselves from talk of conceptual analysis, but even so feel that philosophical reflection is a source of synthetic a priori intuition (Sosa 1998, 2007). They too would seem to face the traditional query of how synthetic a priori knowledge is possible.

In this context, Timothy Williamson has recently argued that the traditional distinction between a priori and a posteriori knowledge is less than clear-cut, and in particular that it breaks down in connection with the intuitions on which philosophers rely. In Williamson’s view, there is a distinctive philosophical method in which intuitive judgements play a central role, but there is no warrant for classifying the relevant intuitions as a priori rather than a posteriori (Williamson 2013).

However, it is arguable that this does not so much address as by-pass the underlying question. Perhaps philosophical intuitions are not best classified as clearly “a priori”. But, if philosophy’s distinctive methodology relies on synthetic intuitions, this still seems to call for some explanation of the source of their reliability.

Doubts about the reliability of philosophical intuitions have been amplified over the past few years by the findings of “experimental philosophy”. Empirical studies have indicated that many central philosophical intuitions are by no means universal, but rather peculiar to certain cultures, social classes and genders (Knobe and Nichols 2008). This variability of intuitions is in obvious tension with their reliability. If different people have opposed philosophical intuitions, then it cannot be that intuitions of this kind are always true.

Timothy Williamson has responded to this challenge from experimental philosophy by suggesting that, while the intuitions of ordinary people on philosophical matters might be unreliable, those of philosophers in particular can be trusted. In his view, a proper philosophical training winnows out mistaken philosophical reactions (2007: 191; 2011). However, this position still seems to call for a positive explanation of how synthetic philosophical knowledge might be established without a posteriori evidence, even if it is restricted to trained philosophers.

The possibility of such an explanation is not of course to be dismissed out of hand. We will do well to remember that a priori does not just mean analytic. There is no contradiction in the idea of experience-independent access to synthetic truths. After all, until the eighteenth century no modern philosopher doubted that God had bestowed on us powers of reason that would enable us to arrive at perception-independent knowledge of a range of synthetic claims. Even if few contemporary philosophers would still appeal to God in this context, there are other possible mechanisms that could play an equivalent role. It is not to be ruled out that our biological heritage, for example, has fixed a range of beliefs in us whose reliability owes nothing to our individual ontogenetic experience. Indeed there is a case for viewing certain aspects of our cultural heritage as playing a similar role, imbuing us with certain beliefs whose justification rests on ancestral rather than individual experience.[19]

However, it is one thing to point to the general possibility of biological and cultural mechanisms constituting experience-independent sources of reliable knowledge, another to show that such mechanisms operate within philosophy. Even if there are areas of thought which rest on such foundations, this does not show that the intuitions of philosophers in particular have a similar backing. Moreover, there is arguably direct reason to doubt that they do. It is not just the intuitions of everyday people on philosophical matter that have a poor track record. The same applies to philosophers through history. It is not hard to think of deep-seated intuitions appealed to by past philosophers that have since been discredited. (A purely mechanical being cannot reason; space must be Euclidean; an effect cannot be greater than its cause; every event is determined; temporal succession cannot be relative.) In the next section we will see that there is reason to suppose that this unreliability is intrinsic to the nature of philosophy.

2.5 The Role of Thought Experiments

A central feature of contemporary philosophy is the use of intuitive judgements about test cases to test philosophical theories. The description theory of names is challenged by Kripke’s imagined counterexamples, the tripartite theory of knowledge by Gettier cases, and so on.

At first pass, this feature of philosophical method might seem to presuppose that philosophy is centrally concerned with analytic claims. After all, isn’t the point of imagined possible cases precisely to show us that certain scenarios are not ruled out by the structure of our concepts, and thereby that certain philosophical theses are not analytically guaranteed?

Because of this, some naturalist philosophers have advocated a revisionary attitude to philosophical method. Philosophy should turn away from the method of possible cases, and instead engage directly with proper empirical theorizing. The interesting questions are not about our concepts of names, knowledge, persons and so on, but about the worldly items, if any, that these concepts refer to (Kornblith 2002; Knobe and Nichols 2008).

However, it is not to be taken for granted that philosophical thought experiments are solely of conceptual significance. Perhaps the intuitions about possible cases are not just analytic consequences of concepts, but substantial synthetic claims about the world. If this were right, then they would be capable of challenging the kind of synthetic theses that methodological naturalists take philosophy to involve.

In support of this alternative, note that intuition arguably plays a role in science as well as philosophy, in the form of scientific thought experiments, like Galileo’s analysis of free fall, or the Einstein’s argument against the completeness of quantum mechanics. Here too the scientist imagines some possible situation, and then makes an intuitive judgement about what would happen. But here the theory at issue is not some conceptual claim, but rather an eminently synthetic one—say, that heavier bodies fall faster. If intuition is to falsify this, then it needs to tell us that there is a naturally and not just conceptually possible situation that violates this thesis—for example, if a big and small body are tied together, they will be heavier than the big one, but will not fall faster. This thought is clearly not guaranteed by concepts alone, but by synthetic assumptions about the way the world works.

But this analogy[20] now highlights a different worry about philosophical thought experiments. If the intuitions promoted by the thought experiments are synthetic, what reason is there to trust them? As was observed in the last section, there is no obvious reason to suppose that the synthetic intuitions of philosophers are generally reliable, and indeed some historical basis for thinking they are prone to error. Given this, it scarcely seems sensible to assess philosophical theories by their consistency with the immediate reactions of philosophers to test cases. Once more, methodological naturalism would seem to require a revisionary attitude to philosophical method.

However, there is a way of understanding the role of thought experiments in philosophy that evades this worry. Instead of viewing them as designed to elicit authoritative intuitions to which philosophical theories must defer, they can instead be seen as devices which help us to articulate our implicit assumptions when we are threatened with paradox and have difficulty finding a solution.

Recall a point made at the beginning of our discussion of methodological naturalism. Philosophical problems are typically occasioned by some kind of theoretical tangle. Different but equally plausible lines of though lead us to conflicting conclusions. Unraveling this tangle requires that we lay out different theoretical commitments and see what might be rejected or modified. A useful heuristic for this purpose may well be to use intuitions about imaginary cases to uncover the implicit synthetic assumptions that are shaping our thinking. This can help us better to appreciate our overall theoretical alternatives, and assess which gives the best overall fit with the a posteriori evidence.

This perspective on philosophical thought experiments shows why we should positively expect many of the intuitions they elicit to prove wanting. Perhaps there are contexts outside philosophy where various kinds of a priori intuitions can be relied upon. But if philosophical problems typically arise because we are unsure about what exactly is amiss in the overall set of synthetic claims we bring to the world, then it would seem only to be expected that the fault will often lie in the implicit assumptions behind our thought-experimental reactions.

It is worth noting that this often happens with scientific thought experiments too. Galileo’s intuition that light bodies fall as fast as heavy ones was vindicated by subsequent physics. But the verdict can also go the other way. For example, the assumption behind the Einstein’s argument against the completeness of quantum mechanics is nowadays rejected. But this certainly did not mean that his thought experiment was worthless. On the contrary, it led J. S. Bell to the derivation of the eponymous inequality whose experimental confirmation ruled out local hidden variable theories.

It is not hard to think of similar philosophical cases. The worth of philosophical thought-experiments does not always require that the intuitions they elicit are sound. Consider the classic Lockean set-up where someone’s memories are transferred to a new body. We all have an intuition that the person goes with the memories, not the old body, as evidenced by our reactions to the many fictions which trade on just this kind of scenario. But few philosophers of personal identity would nowadays hold that this intuition is decisive in favour of Lockeanism. Again, consider the intuition that conscious properties are ontologically distinct from physical ones, as displayed in our immediate reaction to zombie scenarios. Here too, few would suppose that these intuitions by themselves decide the case. Still, even those who reject Lockeanism and dualism will allow that reflection on memory-switching and zombie cases has played a crucial role in clarifying what is at issue in the debates. The evocation of intuitions by philosophical thought experiments is important, not because they provide some special kind of a priori evidence, but simply because they need to be made explicit and assessed against the overall a posteriori evidence.

This perspective on thought experiments shows that there is a sense in which recent developments within “experimental philosophy” can be viewed as complementing traditional armchair methods. In the previous section we saw that some of the findings of experimental philosophy carry the implication that everyday intuitions are not generally be reliable. But in addition to this “negative” message, there is also room for experimental philosophy to make a positive philosophical contribution, even in cases where there is no variation in intuitions.

Careful experimental probing can helpfully augment traditional armchair methods as a way of identifying the structure of implicit assumptions that drive intuitive judgments about test cases. Sometimes thought-experiments may be enough. But in more complicated cases systematic questionnaires and surveys may well be a better way of identifying the implicit cognitive structures behind our philosophical reactions.

Note that experimental philosophy, even when viewed in this positive light, is at most an addition to our philosophical armoury, not a new way of doing philosophy. For once we have sorted out the intuitive principles behind our philosophical judgements, whether by armchair reflection or empirical surveys, we still need to assess their worth. Even if the claim that we think a certain way is supported by hard empirical data, this doesn’t make that way of thinking correct. That can only be shown by subjecting that way of thinking itself to proper a posteriori evaluation.

2.6 Mathematical, Modal and Moral Knowledge

Methodological naturalism fits more naturally with some areas of philosophy than others. It is perhaps not hard to understand, at least in outline, how work in areas like metaphysics, philosophy of mind, meta-ethics and epistemology might be aimed at the construction of synthetic theories supported by a posteriori evidence. But in other philosophical areas the methodologically naturalistic project may seem less obviously applicable. In particular it might be unclear how it applies to those areas of philosophy that make claims about mathematics, first-order morality or modality.

One possibility would be for methodological naturalists to make exceptions for these areas of philosophy. It would still be a significant thesis if methodological naturalism could be shown to apply to most central areas of philosophy, even if some specialist areas call for a different methodology.

This final subsection will address two issues raised by this suggestion. First, is it reasonable to regard modal claims in particular as comprising a specialism within philosophy? Perhaps mathematical investigation and even first-order moralizing can be regarded as marginal to the core areas of philosophy. But many will feel that a concern with the modal realm is a characteristic feature of all philosophy. Second, how far do methodological naturalists need to allow that mathematics, morality and modality constitute exceptions to their position in the first place?

On the first issue, Bertrand Russell said

[A philosophical proposition] must not deal specially with things on the surface of the earth, or with the solar system, or with any other portion of space and time. … A philosophical proposition must be applicable to everything that exists or may exist. (1917: 110)

However, one can agree with Russell that philosophy automatically has implications for the modal realm (“everything that … may exist”), without accepting that the aim of philosophy is to explore the modal realm as such. We need to distinguish here between an interest in claims which, as it happens, have modal implications, on the one hand, and an interest in those modal implications themselves, on the other. It is uncontentious that most of the claims of interest to philosopher have modal implications. But it does not follow from this that most of philosophy is interested in the modal realm itself.

Philosophy is largely concerned with claims about identity and constitution, claims which as it happens will be necessary if they are true. When philosophers ask about knowledge, names, persons, persisting objects, free will, causation, and so on, they are seeking to understand the identity or constitution of these kinds. They want to know whether knowledge is the same as true justified belief, whether persisting objects are composed of temporal parts, and so on. And so any truths they might establish about such matters will inevitably be necessary rather than contingent, and so carry implications about a realm beyond the actual.

But the fact that p implies necessarily p does not mean that anyone who is interested in the former must be interested in the latter, any more than someone who is interested in John’s age being 47 must be interested in its being a prime number.

This makes room for methodological naturalists to insist that most primary philosophical concerns are synthetic and a posteriori, even if they imply additional modal claims which are not. Natural science provides a good analogy here. Water is H2O. Heat is molecular motion. Stars are made of hot gas. Halley’s comet is made of rock and ice. Since all these claims concern matters of identity and constitution, they too are necessary if true. But science is interested in these synthetic a posteriori claims as such, rather than their modal implications. Chemistry is interested in the composition of actual water, and not with what happens in other possible worlds. Methodological naturalists can take the same line with philosophical claims. Their focus is on whether knowledge is actually the same as true justified belief, or whether persisting objects are actually composed of temporal parts—issues which they take to be synthetic and a posteriori—and not with whether these truths are necessary—issues which may well have a different status.

Let us now turn to the second issue flagged above. How far do methodological naturalists in fact need to allow that modality—and mathematics and first-order morality—do have a different status from the synthetic a posteriori character they attribute to philosophy in general?

The issues here are by no means clear-cut. In sections 1.7 and 1.8 above we saw how the arguments for ontological naturalism placed general constraints on the epistemological options in these area. There is no question of exploring these epistemological issues fully here, but some brief comments will be in order.

For mathematics and modality, the epistemological possibilities were restricted to irrealism and ontologically non-naturalist realism. In the moral case, there were again irrealist options, and also ontologically naturalist realisms that identified moral facts with causally significant spatiotemporal facts.

For those who endorse irrealist options in any of these areas, there would seem to be no tension with methodological naturalism. After all, irrealist analyses deny that there is any substantial knowledge to be had in mathematics, modality or morality, and so will not think of object-level claims in these areas as themselves contributing to philosophy. (This is consistent with thinking that a meta-understanding of the workings of mathematical, modal or moral discourse is important to philosophy; but then there seems no reason why such a meta-understanding should be problematic for methodological naturalism.)

Similarly, there seems no reason why the naturalist realist options in the moral case should be in tension with methodological naturalism. The details deserve to be worked through, but on the face of things we might expect knowledge of causally significant spatiotemporal moral facts to be synthetic and a posteriori.

This leaves us with non-naturalist realist accounts of mathematical and modal knowledge. As we saw earlier, the best options here appeal to the neo-Fregean programme of grounding knowledge of the mathematical and modal realms in a priori analytic principles. If this programme could be vindicated, then it would indeed violate the requirements of methodological naturalism. But, as observed earlier, it seems at best an open question whether analytic principles have the power to take us to realist knowledge of the mathematical and modal realms.


  • Armstrong, D., 1968, A Materialist Theory of the Mind, London: Routledge and Kegan Paul.
  • Baker, A., 2005, “Are there Genuine Mathematical Explanations of Physical Phenomena?”, Mind, 114: 223–38.
  • Batterman, R., 2010, “On the Explanatory Role of Mathematics in Empirical Science”, British Journal for the Philosophy of Science, 61: 1–25.
  • Bennett, K., 2003, “Why the Exclusion Problem Seems Intractable, and How, Just Maybe, to Tract It”, Nous, 37: 471–97.
  • Bickle, J., 2013, “Multiple Realizability”, The Stanford Encyclopedia of Philosophy (Spring 2013 edition), Edward N. Zalta (ed.), URL = <>.
  • Blackburn, S., 1993, Essays in Quasi-Realism, Oxford: Oxford University Press.
  • Boyd, R., 1988, “How to Be a Moral Realist”, in G. Sayre-McCord (ed.), Essays on Moral Realism, Ithaca: Cornell University Press, 181–228.
  • Brandom, R., 2001, “Reason, Expression, and the Philosophical Enterprise”, in C. Ragland and S. Heidt (eds.), What Is Philosophy?, New Haven: Yale University Press.
  • Burge, T., 1993a, “Mind-Body Causation and Explanatory Practice”, in J. Heil and A. Mele (eds.), Mental Causation, Oxford: Clarendon Press.
  • –––, 1993b, “Content Preservation”, Philosophical Review, 102: 457–88.
  • Butterfield, J., 2011, “Less is Different: Emergence and Reduction Reconciled”, Foundations of Physics, 41(6): 1065–1135.
  • Cappelen, H., 2012, Philosophy without Intuitions, Oxford: Oxford University Press.
  • Chalmers, D., 1996, The Conscious Mind, Oxford: Oxford University Press.
  • Craig, W. and J. Moreland (eds.), 2000, Naturalism: A Critical Analysis, London: Routledge.
  • Daly, C. and S. Langford, 2009, “Mathematical Explanation and Indispensability Arguments”, Philosophical Quarterly, 59: 641–658.
  • Davidson, D., 1970, “Mental Events”, in L. Foster and J. Swanson (eds.), Experience and Theory, London: Duckworth. Reprinted in Davidson 1980.
  • –––, 1980, Essays on Actions and Events, Oxford: Clarendon Press.
  • DePaul, M. and W. Ramsey (eds.), 1998, Rethinking Intuition, Lanham, Maryland: Rowman and Littlefield.
  • Draper, P., 2005, “God, Science, and Naturalism”, in W. Wainwright (ed.), The Oxford Handbook of Philosophy of Religion, Oxford: Oxford University Press.
  • Elkana, Y., 1974, The Discovery of the Conservation of Energy, London: Hutchinson.
  • Feigl, H., 1958, “The ‘Mental’ and the ‘Physical’”, in H. Feigl, M. Scriven and G. Maxwell (eds.), Minnesota Studies in the Philosophy of Science, Volume II, Minneapolis: University of Minnesota Press.
  • Field, H., 1980, Science Without Numbers, Oxford: Blackwell.
  • –––, 1989, Realism, Mathematics and Modality, Oxford: Blackwell.
  • Fodor, J., 1974, “Special Sciences or: The Disunity of Science as a Working Hypothesis”, Synthese, 28: 97–115.
  • Gibbard, A., 2003, Thinking How To Live, Cambridge, MA: Harvard University Press.
  • Goldman, A. 2007: “Philosophical Intuitions: Their Target, Their Source, and Their Epistemic Status”, Grazer Philosophische Studien, 74: 1–26.
  • Hale, B. and C. Wright, 2003, The Reason’s Proper Study: Essays towards a Neo-Fregean Philosophy of Mathematics, Oxford: Oxford University Press.
  • Hare, R., 1952, The Language of Morals, Oxford: Clarendon Press
  • Harman, G., 1986, “Moral Explanations of Natural Facts”, Southern Journal of Philosophy, 24: 69–78.
  • Hempel, C., 1969, “Reduction: Ontological and Linguistic Facets”, in S. Morgenbesser et al. (eds.), Essays in Honor of Ernest Nagel, New York: St Martin’s Press.
  • Hill, C., 1991, Sensations, Cambridge: Cambridge University Press.
  • Hill, C. and B. McLaughlin, 1999, “There are Fewer Things in Reality than are Dreamt of in Chalmers’ Philosophy”, Philosophy and Phenomenological Research, 59: 445–54.
  • Honderich, T., 1982, “The Argument for Anomalous Monism”, Analysis, 42: 59–64.
  • Hornsby, J., 1997, Simple-Mindedness: In Defence of Naïve Naturalism in the Philosophy of Mind, Cambridge, MA: Harvard University Press.
  • Jackson, F., 1982, “Epiphenomenal Qualia”, Philosophical Quarterly, 32: 127–36.
  • –––, 1986, “What Mary Didn’t Know”, Journal of Philosophy, 83: 291–5.
  • –––, 1993, “Armchair Metaphysics”, in J. O’Leary-Hawthorne and M. Michael (eds.), Philosophy in Mind, Dordrecht: Kluwer.
  • –––, 1998, From Metaphysics to Ethics, Oxford: Clarendon Press.
  • –––, 2003, “Mind and Illusion”, in A. O’Hear (ed.), Minds and Persons, Cambridge: Cambridge University Press.
  • Jenkins, C., 2008, Grounding Concepts: An Empirical Basis for Arithmetical Knowledge, Oxford: Oxford University Press.
  • –––, 2012, “A Priori Knowledge: The Conceptual Approach”, in A. Cullison (ed.), The Continuum Companion to Epistemology, London: Continuum Press.
  • Joyce, R., 2015, “Moral Anti-Realism”, The Stanford Encyclopedia of Philosophy (Summer 2015 edition), Edward N. Zalta (ed.), URL = <>.
  • Kalderon, M., 2005, Moral Fictionalism, Oxford: Oxford University Press.
  • Kim, J., 1998, Mind in a Physical World, Cambridge, MA: MIT Press.
  • –––, 2003, “The American Origins of Philosophical Naturalism”, Journal of Philosophical Research, APA Centennial Volume: 83–98.
  • Knobe J. and S. Nicholls (eds.), 2008, Experimental Philosophy, New York: Oxford University Press.
  • Kornblith, H., 2002, Knowledge and Its Place in Nature, Oxford: Oxford University Press.
  • Krikorian, Y. (ed.), 1944, Naturalism and the Human Spirit, New York: Columbia University Press.
  • Leng, M., 2013, Mathematics and Reality, Oxford: Oxford University Press.
  • Lenman, J., 2014, “Moral Naturalism”, The Stanford Encyclopedia of Philosophy (Spring 2014 edition), Edward N. Zalta (ed.), URL = <>.
  • Lewis, D., 1966, “An Argument for the Identity Theory”, Journal of Philosophy, 63: 17–25.
  • –––, 1970, “How to Define Theoretical Terms”, Journal of Philosophy, 67: 427–46.
  • List, C. and P. Menzies, 2009, “Non-Reductive Physicalism and the Limits of the Exclusion Principle”, Journal of Philosophy, 106: 475–502.
  • –––, 2010, “The Causal Autonomy of the Special Sciences”, in C. Macdonald and G. Macdonald (eds.), Emergence in Mind, Oxford: Oxford University Press.
  • Loewer, B., 2007, “Counterfactuals and the Second Law”, in H. Price and R. Corry (eds.), Causation, Physics, and the Constitution of Reality, Oxford: Oxford University Press.
  • Lowe, E.J., 2000, “Causal Closure Principles and Emergentism”, Philosophy, 75: 571–85.
  • –––, 2003, “Physical Causal Closure and the Invisibility of Mental Causation”, in S. Walter and H.-D. Heckmann (eds.), Physicalism and Mental Causation, Exeter: Imprint Academic.
  • MacBride, F., 2003, “Speaking with Shadows: A Study of Neo-Logicism”, British Journal for the Philosophy of Science, 54: 103–63.
  • McDowell, J., 1996, Mind and World, Cambridge, MA: Harvard University Press.
  • Mackie, J., 1977, Ethics: Inventing Right and Wrong, Harmondsworth: Penguin.
  • Maddy, P., 1990, Realism in Mathematics, Oxford: Oxford University Press.
  • Mellor D., 1995, The Facts of Causation, London: Routledge.
  • Menon, T. and C. Callender, 2013, “Turn and Face the Strange … Ch-ch-changes”, in R. Batterman (ed.), The Oxford Handbook of Philosophy of Physics, Oxford: Oxford University Press.
  • Menzies, P., 2008, “Causal Exclusion, the Determination Relation, and Contrastive Causation”, in J. Kallestrup and J. Hohwy (eds.), Being Reduced: New Essays on Reductive Explanation and Special Science Causation, Oxford: Oxford University Press
  • Montero, B. and D. Papineau, 2005, “A Defence of the Via Negativa Argument for Physicalism”, Analysis, 65: 233–7.
  • Moore, G., 1903, Principia Ethica, Cambridge: Cambridge University Press.
  • Nagel, J., 2007, “Epistemic Intuitions”, Philosophy Compass, 2/6: 792–819.
  • Nolan, D., 2011a, “Modal Fictionalism”, The Stanford Encyclopedia of Philosophy (Winter 2011 edition), Edward N. Zalta (ed.), URL = <>.
  • –––, 2011b, “Modal Fictionalism and Possible World Semantics” (supplement to “Modal Fictionalism”, Winter 2011 edition), The Stanford Encyclopedia of Philosophy, Edward N. Zalta (ed.), URL = <>.
  • Oppenheim, H. and P. Putnam, 1958, “Unity of Science as a Working Hypothesis”, in H. Feigl, M. Scriven, and G. Maxwell (eds.), Minnesota Studies in the Philosophy of Science, Volume II, Minneapolis: University of Minnesota Press.
  • Papineau, D., 2002, Thinking about Consciousness, Oxford: Oxford University Press.
  • –––, 2013, “Causation is Macroscopic but not Irreducible”, in S. Gibb, E. Lowe, and R. Ingthorsson (eds.), Mental Causation and Ontology, Oxford: Oxford University Press.
  • –––, 2014, “The Poverty of Conceptual Analysis”, in M. Haug (ed.), Philosophical Methodology, London: Routledge.
  • –––, 2015, “The Nature of A Priori Intutions: Analytic or Synthetic?”, in E. Fischer and J. Collins (eds.), Experimental Philosophy, Rationalism, and Naturalism, London: Routledge.
  • Parsons, C., 1995, “Platonism and Mathematical Intuition in Kurt Gödel’s Thought”, Bulletin of Symbolic Logic, 1(1): 44–74.
  • Plantinga, A., 1996, “Methodological Naturalism?”, in J. van der Meer (ed.), Facets of Faith and Science, Lanham, Maryland: University Press of America.
  • Price, H., 2011, Naturalism without Mirrors, Oxford: Oxford University Press.
  • Pust, J., 2014, “Intuition”, Stanford Encyclopedia of Philosophy (Fall 2014 edition), Edward N. Zalta (ed.), URL = <>.
  • Putnam, H., 1971, Philosophy of Logic, New York: Harper.
  • Ridge, M., 2014, “Moral Non-Naturalism”,

A Defense of Naturalism

by Keith Augustine

Advisory Committee:

Professor Raymond F. Martin, Chair
Professor Allen Stairs
Professor Corey Washington


Chapter 1: Naturalism and the Natural-Supernatural Distinction

What is Naturalism?
The Meaning of 'Nature' or 'Natural'
The Natural-Supernatural Distinction

Chapter 2: An Empirical Case for Naturalism

Extraordinary Claims Require Extraordinary Evidence
Ockham's Razor
Metaphysical and Methodological Naturalism
A Natural History of the Universe
Naturalism and a Scientific Picture of the World
The Status of Parapsychological Research



Throughout the twentieth century naturalism has been a label for a variety of distinct positions which have little, if anything, in common. In ethics, naturalism is a form of moral realism which contends that ethical properties are objective in virtue of being reducible to or identical to natural properties, where natural properties are simply the properties investigated by various sciences. In metaphysics, naturalism typically takes a form of materialism or physicalism: Everything that exists is either physical or supervenient upon the physical. Naturalism in epistemology contends that the role of epistemology is to describe how knowledge is obtained rather than to set out a priori criteria for the justification of beliefs; thereby a naturalized epistemology provides theories of knowledge and justification which eliminate normative standards by using only scientific concepts.

In this essay I will be concerned with naturalism in the philosophy of religion, where other basic metaphysical and epistemological issues will arise. Naturalism in this domain is the antithesis of supernaturalism--it is often construed as the view that everything that exists is natural and thus by implication that the supernatural does not exist. Behind this superficially simple formulation of naturalism, however, lies a wealth of implicit complexity. Part of this complexity consists of the analysis of the meaning of the word 'nature' or 'natural', how nature is to be characterized, and how the natural-supernatural distinction is to be drawn, both in theory and in practice. These issues will be addressed in the first part of this essay. In the second part I will defend naturalism as a more reasonable option for belief than either supernaturalism or agnosticism. This defense of naturalism will rest on an argument which contends that the lack of uncontroversial evidence for potential instances of supernatural causation provides strong inductive grounds for taking naturalism to be true.

What is Naturalism?

One of the most common versions of naturalism is the position that everything that exists is natural. Robert Audi defines naturalism, broadly construed, as "the view that nature is all there is and all basic truths are truths of nature" (Audi 1996, p. 372). Rem B. Edwards offers a similar definition: "[T]he naturalist is one who affirms that only nature exists and by implication that the supernatural does not exist... The [natural] world is all of reality; it is all there is; there is no 'other world' " (Edwards 1972, p. 135). Although these definitions capture some of the most fundamental features of naturalism, I think that naturalism can be--and thus should be--defined less strongly. Alan Lacey captures the heart of naturalism when he writes: "What [naturalism] insists on is that the world of nature should form a single sphere without incursions from outside by souls or spirits, divine or human" (Lacey 1995, p. 604).

I think that most naturalists would agree that naturalism at least entails that nature is a closed system containing only natural causes and their effects. Fundamentally, naturalism is a metaphysical position about what sorts of causal relations exist--it is the position that every caused event within the natural world has a natural cause. This definition of naturalism is weaker than "everything that exists is natural" because it leaves open the possibility that the natural world does not exhaust all of reality: There may be some aspects of reality which exist outside of nature. Which aspects of reality are nonnatural in this sense will vary with the different definitions of nature or natural being used. It may even be impossible in principle to know that such nonnatural realms exist. But this weaker definition retains the fundamental core of naturalism by denying that supernatural causation exists. It would thus be better to say that naturalism is the position that everything that exists within nature is itself natural and is solely influenced by natural causes.

Naturalism, as I conceive it, thus allows the existence of both nature and realms that may exist outside of nature; it simply stipulates that any nonnatural realms which may exist cannot causally influence the natural world. Even the possibility of nonnatural causation is not ruled out so long as both the cause and effect reside in some nonnatural realm. Thus naturalism allows for the existence of both the natural and the nonnatural--including instances of natural and nonnatural causation--as long as these domains are causally separate. A supernatural cause, on this view, would be a nonnatural cause of an event within nature. The phrase 'supernatural event' is best taken to refer to an event within nature which has a supernatural cause. The phrase 'natural event' can refer to either an event with a natural cause or an event in the natural world. We should distinguish between these two, so I will not use the phrase 'natural event'. Instead, I will use the phrases 'naturally-caused event' and 'event within nature' (or the natural world), respectively, to mark this distinction. Naturalism is thus best construed as the denial of the existence of any genuine instances of supernatural causation, whereas supernaturalism is the affirmation of the existence of such instances.

Arthur C. Danto comes closest to explicitly defining naturalism in this way when he characterizes naturalism as entailing that "The entire knowable universe is composed of natural objects--that is, objects which come into and pass out of existence in consequence of the operation of 'natural causes' " (Danto 1972, p. 448). But what is a natural cause? According to Danto,

A natural cause is a natural object or an episode in the history of a natural object which brings about a change in some other natural object... [I]t is solely with reference to natural causes that we explain changes in the behavior of natural objects. This may require reference to objects which we cannot directly experience, but these will nevertheless still be natural objects, and we need never go outside the system of natural objects for explanations of what takes place within it. Reference to nonnatural objects is never explanatory (Danto 1972, p. 448).

Insofar as the meaning of the term 'natural' is not made explicit, the definition above leaves open the possibility that 'natural cause' might be defined broadly as any cause of a change in the behavior of a natural object. Such a broad definition of 'natural cause' clearly begs the question: That all causes of events within nature are natural causes is precisely the issue in question. We certainly don't want this thesis to be true by definition--that is, true in a trivial sense. Rather, we want naturalism to be a position which--if true--is informative. The poignant feature of Danto's definition which seems most essential to naturalism is the thesis that we never need to look to something outside of the natural world to explain anything within the natural world.

On Danto's definition, we may not always be able to directly experience a natural cause, but presumably we should be able to experience it indirectly, as when we think of atoms as natural objects. While Danto never states how he distinguishes between directly experiencing an object and indirectly experiencing it, I will presume that he means something like the following: An object is directly experienced if it is immediately present to our senses; it is indirectly experienced if we must infer its presence to explain the behavior of other objects which are immediately present to our senses[1]. Danto's discussion of nonnatural objects indicates that he does not intend 'natural cause' to refer simply to any cause of a change in a natural object:

The universe may in addition contain one or another sort of nonnatural object, but we have no reason for allowing the existence of nonnatural objects unless they have impact on the observable behavior of natural objects, for natural objects are the only objects about which we know directly, and it would be only with reference to their perturbations that we might secure indirect knowledge of nonnatural objects, should there be any (Danto 1972, p. 448).

Suppose we grant Danto his assumption that only natural objects can be known directly[2]. A crucial question still arises: Among indirectly-known objects, how do we distinguish between those which are natural and those which are nonnatural?

The Meaning of 'Nature' or 'Natural'

Danto's definition of a natural cause, while capturing very general features of natural causation and natural causal explanation, does not shed much light on what is meant by the term 'natural' itself. One obvious candidate for what is meant by the term 'natural' is physical. The earliest forms of naturalism, in fact, were versions of materialism or physicalism which maintained that everything that exists is physical. As I have construed naturalism, simple (reductive) physicalism maintains that everything that exists within nature is physical and solely influenced by physical causes. However, the prominent twentieth century debate over materialism in the philosophy of mind has revealed several difficulties with reductive physicalism as a solution to the mind-body problem.

One of the most persistent difficulties for reductive physicalism has been the apparent inability of physicalistic explanations to capture qualitative features of conscious experience. It has been persuasively argued that qualia--the experiential feels of 'what it is like' to be in a conscious mental state--cannot be captured by any physicalistic explanations in principle because physicalistic explanations inherently refer to objective or public features of phenomena, whereas the experiential features of consciousness are inherently subjective or private (Teller 1992, pp. 190-191). While such arguments for the irreducibility of consciousness are not the last word on the subject, they have not been decisively refuted either--at least not in the view of several prominent philosophers. Although such difficulties may be resolved in the future, their current resistance to a clear resolution that gains widespread acceptance gives us good reason to resist simply identifying the natural with the physical.

In the contemporary philosophy of mind, an attractive alternative to reductive physicalism is some version of nonreductive physicalism or property dualism[3]. According to nonreductive physicalism, mental states are not simply identical to certain physical states (such as brain states), as reductive physicalists hold; rather, mental states are supervenient upon those physical states. There have been several competing definitions of supervenience suggested in the philosophical literature[4]. In general, however, to say that mental states supervene upon physical states is to say that there can be no differences between mental states without a physical difference between the objects which instantiate those states (Beckermann 1992, p. 11). This physical difference usually amounts to a difference in brain states, though the same mental states may be supervenient upon the physical states of an advanced computer or of an extraterrestrial brain. For our purposes, it is sufficient to say that for a mental state to be supervenient upon a physical state entails that a mental state is dependent upon and determined by that physical state without necessarily being identical to it.

But if mental states are supervenient upon some physical states and are not identical to any physical states, this means that mental states are--by definition--nonphysical. If we accept nonreductive physicalism (or even admit it as a reasonable position) and want to retain naturalism, we do not want to say that 'natural' is simply equivalent to 'physical'. However, the driving idea behind nonreductive physicalism allows us to consider another candidate for the natural: perhaps the term 'natural' means physical or supervenient upon the physical[5]. On my definition of naturalism, nonreductive physicalism maintains that everything that exists within nature is either physical or supervenient upon the physical and solely influenced by physical causes or causes which are supervenient upon physical causes. A more economical statement of this form of naturalism would drop the idea of supervenient causation: everything that exists within nature is either physical or supervenient upon the physical and solely influenced by physical causes. Most reductive and nonreductive physicalists alike subscribe to the causal closure of the physical--the view that all caused events in the physical world must have physical causes (Van Gulick 1992, p. 160). Moreover, nonphysical causation is unlikely given that the brain would behave noticeably differently under the constant influence of nonphysical causes than it would in the absence of such influence and we see no evidence for nonphysical influences on the brain.

If naturalism is construed as the position that everything that exists is natural, the definition of natural as 'physical or supervenient upon the physical'--though initially promising--runs into potential difficulties. Consider the philosophical debate over the existence of abstract objects. According to Platonism, there exists a class of mind-independent entities called abstract objects (Hale 1987, p. 11). On traditional Platonic accounts, abstract objects are immutable and timeless entities which are incapable of being involved in causal interactions--that is, are acausal--because they exist outside of space and time in a Platonic realm of unchanging and eternal forms. A paradigm candidate for a genuine abstract object is a number:

Numbers, sets and other stock examples of the abstract have neither spatial nor temporal position. Someone who seriously persisted in asking after the whereabouts of the number 3, say, or when it began to exist, or how long it will endure, etc., could only be supposed to be the victim of a gross misconception concerning what kind of thing numbers are (or are taken to be). With such paradigmatic examples of the abstract in mind, it is natural to propose that the distinguishing feature of abstract objects is lack of spatial or temporal location (Hale 1987, p. 48).

However, it is questionable whether Platonism must be characterized in this way. For example, Bob Hale points out that while all candidates for abstract objects are nonspatial, certain candidates for abstract objects, such as the game of chess and the English language, have an origin in time (Hale 1987, p. 49). One could argue that such examples are not genuine abstract objects after all, though Hale thinks that this is implausible. Despite this assessment, however, Hale does concede that "the vast majority of abstract objects surely are wholly atemporal as well as non-spatial" (Hale 1987, p. 253). Perhaps the only abstract objects which we are forced to countenance as real, if we are forced to countenance any at all, are those which clearly exist outside of space and time. This would explain why abstract objects are in some sense acausal. Hale points out that while it isn't obvious that abstract objects must be completely acausal, "when abstract objects are said to be constitutionally incapable of causal involvement, what is meant is that they cannot be causes of change, and perhaps also that they cannot undergo change" (Hale 1987, p. 2). Given Danto's understanding of a cause as something "which brings about a change" in an object, abstract objects are acausal in the sense of causality that we are interested in.

In any case, I will confine our exploration of the controversy over abstract objects to paradigm cases of abstract objects like numbers where the traditional definition of abstract objects does apply. There is nothing we can point to within space and time and say 'that is the number 4'. Furthermore, numbers and the relations between them are unchanging and mathematical truths like 2+2=4 seem timelessly true. Physical objects such as acorns can be arranged such that we can say that there are only four of those objects within a given space, but these objects exemplify instances of the number 4--they are not equivalent to '4' itself. On a Platonic account, four acorns are a concrete and particular exemplification of this abstract and universal form. So 4 is a universal concept rather than a particular one. The number 4 is also an abstract concept rather than a concrete one, unlike the idea of an acorn. We cannot point to the number 4 in the way we can point to an acorn--this is the essence of what being an abstract object is.

Does naturalism allow the existence of abstract objects? Alan Lacey thinks that naturalism construes the natural world as a closed system of natural causes and effects "without having to accommodate strange entities like non-natural values or substantive abstract universals" (Lacey 1995, p. 604). Similarly, Arthur C. Danto thinks that naturalism entails the denial of the existence of abstract objects. Danto argues that formal sciences like mathematics

no more entail a Platonistic ontology than [the empirical sciences do], nor are we, in using algorithms, committed to the existence of numerical entities as nonnatural objects. If the formal sciences are about anything, it will at least not be a realm of timeless numerical essences, and at any rate logic and mathematics are properly appreciated in terms not of subject matter but of function, as instruments for coping with this world rather than as descriptions of another one (Danto 1972, p. 449).

Robert Audi, by contrast, thinks that naturalists can admit the existence of abstract objects, noting that they would still be naturalists 'about the world': "[A] naturalist does not have to be a radical physicalist--taking the position that only physical phenomena are real, not even excepting such well-behaved abstract entities as sets" (Audi 2000, p. 31). Audi argues that abstract objects may be essential for any adequate ontology: "It is even more obvious that it could turn out to be impossible to give an adequate account of science, not to mention philosophy, without positing some kinds of abstract entities, such as numbers, propositions, and possible worlds" (Audi 2000, p. 32).

What grounds do we have to believe that abstract objects actually exist? There has been a persuasive argument that at least some abstract entities, particularly mathematical objects such as numbers and sets, are indispensable to our best scientific theories (Hale 1987, p. 104). Insofar as our best scientific theories rely on mathematical descriptions which (allegedly) presuppose the existence of mathematical abstract objects and we ought to believe those theories, we ought to (so the argument goes) accept the existence of at least some abstract objects. The grounds for belief in abstract objects, on this view, are equivalent to the grounds we have for believing in the existence of theoretical entities essential to physics, such as electrons. Of course there are also grounds for doubting the existence of any abstract objects. A common nominalist objection is that by their acausal nature abstract objects are impossible to detect (directly or indirectly) in principle and thus it would be impossible for us to have knowledge of them even if they did exist (Hale 1987, p. 79). Many nominalists conclude that since Platonists do claim to have knowledge of abstract objects and such knowledge is impossible in principle, the concept of an abstract object is incoherent and thus abstract objects do not exist.

At this point in history there is no clear resolution to the question of whether abstract objects exist. However, many philosophers believe that we must admit the existence of at least some Platonic abstract objects if we are going to provide an adequate account of the world. Because intuitively, at least, there do seem to be abstract objects which are acausal and exist outside of space and time, we have grounds for thinking that the 'physical or supervenient upon the physical' criterion for what it means to be natural may be deficient. Such a criterion will certainly not allow for the existence of natural abstract objects. While it is still an open question whether such objects actually exist, if we do want to admit the possibility of their existence on this criterion we would have to admit that such objects (if real) are nonnatural.

Because abstract objects are acausal, at least in the sense of not being causes of change, they cannot be supernatural. But depending on how strongly you characterize naturalism, abstract objects may or may not count as part of reality. If naturalism means that everything is natural and natural means "physical or supervenient upon the physical", then naturalism entails a denial of the existence of abstract objects since abstract objects are neither physical nor supervenient upon anything physical. However, if naturalism merely entails that everything that exists within nature is itself natural and is solely influenced by natural causes, then naturalism can admit the existence of nonnatural abstract objects which exist outside of nature and do not causally influence the natural world. Arguably, the issues involved in the Platonic realism-nominalism debate[6] should be settled on grounds independent of the truth or falsity of naturalism or how naturalism is characterized.

Even if we accept a weaker definition of naturalism which allows for the existence of nonnatural abstract objects and we want to say that 'natural' is equivalent to 'physical or supervenient upon the physical', it isn't clear what 'physical' itself means. In order to determine what 'physical' means it seems that we have to identify characteristics that all physical things have in common--that is, we have to identify fundamental physical properties. Anything belonging to the category 'physical' would at least exhibit those properties. Identifying properties common to all physical objects does not seem too daunting a task: All physical objects seem to have mass, for example. But physical objects do not exhaust the category 'physical'; the category also includes forms of energy, types of events, physical processes, and even spacetime itself. William P. Alston points out how difficult it is to determine what it means for something to be physical:

I take it to be reasonably clear what it is for a substance to be a material substance. Being spatially extended (plus, perhaps, having certain fundamental physical properties like mass) would seem to be necessary and sufficient. But what it is to be a material (physical) state, property, process, or event presents considerably more difficulty. States and properties (and perhaps events and processes as well) are not susceptible to spatial extension in the clear, unproblematic way substances are... [W]e might say that a state, event, or process is physical if it is definable as the exemplification (by a physical substance?) of one or more physical properties (Alston 2000).

But if we label something physical because it exemplifies physical properties we are back to where we started--we still have not said what it means for a property to be physical. While we can list several clear-cut examples of physical properties (e.g. mass, electric charge, gravitational attraction), we cannot say what it is about those properties that makes them physical.

The possibility of property dualism as a solution to the mind-body problem also makes it impossible to identify physical properties simply as the properties of physical substances--arguably, some physical substances (such as functioning brains) will have nonphysical mental properties. Alston sees the debate between reductive physicalists and property dualists as evidence of a lack of any clear notion of the physical:

What is our concept of 'physical state' such that it is a sensible question as to whether my believing that p or my being visually presented with a maple tree is or is not a physical state, given that I do not recognize the possibility of a non-physical substance of which it is a state? Spatial location cannot be the issue. States of me are not spatially located in the way in which I am. Again, it may be said that the question is as to whether the property of believing that p is a physical property. But what is that issue? What does it take to make a property physical? Presumably being a property of a physical substance is not sufficient, or there would be no controversy here. So what are the pure materialist and the property dualist arguing about? (Alston 2000).

Alston also points out that physical forces have physical properties yet do not necessarily belong to an extended physical substance.

This leads us to an alternative possibility: Perhaps we should define 'physical' as that which constitutes the subject matter of the physical sciences. That is, that which is physical is that which physics countenances as part of its subject matter. There are at least two problems with this suggestion, however. First, certain concepts (such as mental categories) may be countenanced into physical science in the future, although they are not now:

If... we do not characterize the physical intrinsically in terms of some kind of property, we appear forced to define it by appeal to what physical scientists discover, or perhaps would ultimately discover. Then we cannot know a priori that, for example, irreducibly mentalistic explanations will not ultimately be part of what the people we call physicists consider their best overall account of reality (Audi 1996, p. 373).

Second, by defining 'physical' as what physicists study, we make an ontological category relative to the state of scientific knowledge:

When we define the physical by reference to what physical scientists think and work with, what period of the history of science are we to pick for this purpose? It seems arbitrary to canonize the present moment as defining a central metaphysical category. But if we make the definition relative to the stage of scientific development we are considering, then the metaphysical issues change with each fundamental shift in science (Alston 2000).

If abstract objects exist we can still maintain that the category 'natural' is equivalent to 'physical or supervenient upon the physical' by arguing that abstract objects are nonnatural. The possible existence of abstract objects only gives us reason to construe naturalism less strongly than it often is, admitting the possible existence of the nonnatural (but not the supernatural). However, as the discussion above illustrates, it is far from clear what 'physical' itself means. The best we may be able to do is say what sorts of things are clearly physical and what sorts of things appear to be nonphysical and admit that we have no clear criteria for distinguishing between the two. Perhaps this is an ambiguity we can live with--after all, as we soon shall see, there is going to be ambiguity in every definition of 'natural' considered. Nevertheless, this gives us some reason to consider other potential candidates for the category 'natural' in the hope that they will be less ambiguous than 'physical or supervenient upon the physical'.

Another candidate for what the category 'natural' refers to is the spatiotemporal--the 'natural' is that which exists within space and time. On this criterion, for an object to be natural it must have spatial extension and temporal duration; thus abstract objects are nonnatural on this criterion as well. Danto claims that "[e]very natural object exists within the spatiotemporal and the causal orders" (Danto 1972, p. 448). Similarly, Rem B. Edwards defines nature as "the spatiotemporal universe as a whole existing independently of [a] knowing mind" (Edwards 1972, p. 135). Edwards also implies that 'natural' is synonymous with 'spatiotemporal' when he claims that naturalism entails the following: "[A]ll natural events have causes[7] that are themselves natural events. The occurrence of every spatiotemporal event is caused by some other spatiotemporal event or events" (Edwards 1972, p. 136). The fundamental dispute between naturalists and supernaturalists, according to Herbert Spiegelberg, is whether the spatiotemporal exhausts all of reality:

Both [naturalists and supernaturalists] seem to understand by 'nature' the sum total of all occurrents in time and space which are (1) explainable by other occurrents within the same order... and which are (2) usually, though perhaps not necessarily, subject to change... The issue between the two camps would thus seem to be whether nature so defined exhausts the content of reality or whether other entities are to be added for a complete account of the universe (Spiegelberg 1951, p. 342).

An immediate concern is that it is not clear how the spatiotemporal differs from the physical. Is the spatiotemporal simply equivalent to the physical? We do not seem to distinguish between the two terms in everyday usage. However, physicists have speculated about the possible existence of physical entities which are not spatiotemporal. For example, cosmologists have postulated the existence of a singularity--a point of infinite density where the known laws of physics break down--at the beginning of the universe and also believe that singularities lie at the center of black holes. According to Kip Thorne, somewhere near the singularity the unknown laws of quantum gravity take over, replacing spacetime with an exotic quantum foam:

Quantum gravity then radically changes the character of spacetime: It ruptures the unification of space and time into spacetime. It unglues space and time from each other, and then destroys time as a concept and destroys the definiteness of space. Time ceases to exist; no longer can we say that 'this thing happens before that one,' because without time, there is no concept of 'before' or 'after.' Space... becomes a random, probabilistic froth (Thorne 1994, pp. 476-77).

Quantum foam is a theoretical instance of something physical though not spatiotemporal. Furthermore, there may be other universes than the one we inhabit. If we presume that different universes are causally isolated from each other and space and time are merely specific instances of dimensions, it is possible that there will be other universes which are not spatiotemporal but have a different sort of dimensional existence which we cannot even begin to fathom. Moreover, if other universes are spatiotemporal, their spatiotemporal framework will not be our spatiotemporal framework--other universes may have different histories, different laws of physics, and different dimensions than our own. Universes may 'bubble off' from a quantum foam in a sort of 'metauniverse' from which all universes originate--but then that metauniverse will not itself be spatiotemporal.

I imagine that our notion of the natural and the physical should be broad enough to encompass singularities, other universes, a possible metauniverse, and other exotic possibilities raised in cosmology. 'Physical' seems to encompass more than the familiar dimensions of space and time and its contents--it will certainly cover any other dimensions that physicists may postulate. Thus the spatiotemporal seems to fall within the larger domain of the physical. If the physical is too restrictive a category to be identified with the natural, and the physical is broader in scope than the spatiotemporal, then the spatiotemporal criterion would also exclude things we want to count as natural. Thus 'natural' should not be identified with 'spatiotemporal'.

Perhaps by 'natural' we should mean something with a more epistemological flavor than the previous criteria we have considered. Danto also characterizes the natural in terms of scientific knowledge rather than solely in terms of ontological status:

Naturalism... is a species of philosophical monism according to which whatever exists or happens is natural in the sense of being susceptible to explanation through methods... paradigmatically exemplified in the natural sciences... Hence, naturalism is polemically defined as repudiating the view that there exists or could exist any entities or events which lie, in principle, beyond the scope of scientific investigation (Danto 1972, p. 448).

On this criterion, the natural is that which is amenable in principle to scientific investigation. Abstract objects, if they exist, are not scientifically accessible in the way that our paradigm case of natural objects--physical objects--are, and thus should fall within the domain of the nonnatural here as well. Note that on this criterion naturalism is still construed as a metaphysical position even though it is cast in epistemological terms: it is the position that everything that exists within nature is amenable in principle to scientific investigation and solely influenced by causes which are amenable in principle to scientific investigation. This invokes metaphysical claims about what exists within nature and what sorts of causes influence events within nature but characterizes its ontology in epistemological terms as what can be explained scientifically. Presumably, something is amenable in principle to scientific investigation in virtue of some more basic property or properties. However, this criterion alone does not spell out what those properties may be. In my final candidate for the 'natural'--the candidate immediately following this one--I will consider such a property. But for now I will construe the 'amenable in principle to scientific investigation' criterion without spelling out what properties something must have to be amenable in this sense.

There are many phenomena which are not amenable to scientific investigation given our current state of knowledge but intuitively we want to say that some phenomena are clearly scientifically explicable in principle even if we cannot explain them given our current state of knowledge. For example, it isn't clear what happens to a piece of matter when it falls into a black hole but we don't want to say that such a question is scientifically inexplicable in principle. It is simply inexplicable given our current state of physical theory. A future theory of quantum gravity may yield confirmed testable predictions which would allow us to answer that question by determining what the theory predicts about such situations.

However, there are other phenomena which do appear to be scientifically inexplicable in principle. Qualia, for example, resist current scientific explanations because they seem to be inherently private or subjective, whereas scientific explanations seem to require, as a matter of principle, explanations in terms of objective or publicly-accessible features of the world. While the qualia problem may ultimately be resolved, perhaps by denying that qualia really are inherently subjective or by denying that scientific explanations must be cast in terms of objective features of the world, at this point there is no obvious solution in sight. Since biological organisms which are clearly part of the natural world are the subjects of qualitative states we should categorize qualia as falling within the domain of the natural. If there can be no scientific explanation of qualia in principle--which is a tenable position--and we want countenance qualia as part of the natural world, we should not simply identify the natural with that which is amenable in principle to scientific investigation.

One might also object that if there are any other universes which are causally isolated from the universe we inhabit, by definition the contents of these universes would be inaccessible to scientific investigation[8]. Yet we would probably want to countenance any other universes, should there be any, as part of the natural realm[9]. This argument is not as persuasive as the qualia example, however, because the contents of causally-isolated universes would not be amenable in principle to scientific investigation only to us, not to any inhabitants of those particular universes. Nevertheless, the possibility of scientifically inexplicable qualia gives us some reason to resist identifying the natural with that which is amenable in principle to scientific investigation.

In addition to this theoretical problem, it is immediately apparent that this criterion is problematic in a fundamental way: by including the qualifier 'in principle' the criterion is too ambiguous, as it stands, to be of much use. Essentially it appears impossible to say a priori whether or not a given phenomenon is amenable in principle to scientific investigation. The best we can do is say that phenomena for which we have successful scientific explanations are definitely amenable in principle to scientific investigation since they are so amenable in practice. But for phenomena which resist scientific explanation, we cannot say categorically that they are or are not scientifically explicable in principle. The only way that we could determine that a phenomenon is scientifically explicable in principle is by providing a successful scientific explanation of that phenomenon. The problem is that this approach provides an a posteriori answer to such questions when we need an a priori criterion for determining whether something is amenable to scientific investigation in principle.

I do not see any way out of this dilemma. This gives us good reason to abandon 'amenable in principle to scientific investigation' as the sole criterion for the natural. However, if we put qualia aside for the moment, there still may be hope for the criterion if we can find some more basic property or properties by which something is amenable to scientific investigation in principle. By spelling out what properties make an object or event amenable in principle to scientific investigation we may be able to avoid this dilemma.

The final criterion for what makes something natural is that it behaves in accordance with laws of nature. Natural causes are clearly amenable to empirical investigation in principle because they exhibit exceptionless or probabilistic regularities which can be described by laws of nature. Laws of nature are factual statements which accurately describe the behavior of natural objects. By definition, natural laws never change and apply to everything within nature. It is the very existence of lawful regularities which allows us to form causal relationships. If laws of nature did not exist, we could not describe some events as causes and others as effects; the concept of natural causation would have no meaning because events would occur randomly without any recognizable pattern. Any phenomenon which is amenable to scientific investigation can be framed in terms of natural causal relationships established by the discovery of lawful regularities. Thus behaving in accordance with natural laws is a more fundamental criterion for the natural than being amenable in principle to scientific investigation: obeying laws of nature is a prerequisite for being amenable to scientific investigation. Naturalism, on this criterion, contends that everything that exists within nature behaves in accordance with laws of nature and is solely influenced by causes which obey natural laws[10].

On this criterion--like all the others we have considered--abstract objects are nonnatural. Anything which does not obey laws of nature is nonnatural. Abstract objects do not obey natural laws because they do not exist within nature--the laws of nature simply do not apply to them. However, the laws of nature do apply to all natural objects and events. An event within the natural world which has a cause that does not obey the laws of nature is an event with a supernatural cause. If we exclude the possibility of causal overdetermination for the sake of explanatory simplicity--that is, we exclude the possibility that a supernatural entity could cause an event which could be explained solely in terms of natural causes--then any event within nature which resulted from a supernatural cause would behave in a way which was not in accordance with the laws of nature.

If natural causes are sufficient to bring about an event then an appeal to supernatural causation to explain such an event is at least prima facie unwarranted since it is unnecessary and uninformative. If we defined supernatural causation in such a way that it fell within the scope of purely natural causation it would be impossible to provide a reliable basis for distinguishing between an event with a natural cause from an event with a supernatural cause. Thus, we should assume that a supernatural event--any event within nature (and thus within the domain of laws of nature) which had a supernatural cause--would be a violation of a law of nature. This retains the fundamental idea of the supernatural as an intervention into nature. A phenomenon which does not involve the violation of laws of nature by definition is a natural phenomenon. A supernatural event--such as the parting of the Red Sea as described in the Old Testament--is still presumably governed by some laws of nature: for example, the water still coheres together and holds most of its properties. However, some nonnatural force seems to counteract the natural forces in place, and this amounts to a violation of a law of nature.

More needs to be said about how natural laws themselves are to be understood. I will distinguish actual laws of nature from scientific laws on the following assumption: Scientific laws (such as laws of physics) are inaccurate but reasonable approximations of genuine laws of nature (Armstrong 1983, p. 6). I will also assume a necessitarian account of laws of nature--an account which contends that the actual laws of nature accurately describe physical necessities in nature, whether those necessities are properties of laws themselves or inhere in the constituents of the universe (Leckey and Bigelow 1995, p. 92). This contrasts with a regularity account of laws of nature, according to which laws of nature simply accurately describe which sorts of events happen in the entire history of the universe (Armstrong 1983, p. 32). The difference between these views is that necessitarian accounts imply that laws of nature 'forbid' certain events from happening--the laws of nature 'dictate' that they cannot happen. Events which cannot happen according to natural laws--that is, violations of laws of nature--are physically impossible events. Regularity accounts of laws of nature deny that there is any sort of physical possibility--there is only the logically possible and the purely contingent. A negation of a natural law is simply a description of a logically possible event that never happens in the entire history of the universe.

A necessitarian might argue that any spacecraft we develop cannot travel faster than the speed of light because that would be a physically impossible event--the laws of nature forbid such an event--it cannot happen. A regularity theorist would argue that such an event simply never occurs in the history of the universe--it simply does not happen. On a necessitarian account of laws of nature, such an event would be physically impossible, but on a regularity account, any logically possible event could happen--it just turns out that in the entire history of the universe certain events do not happen even though they could have happened. A major difficulty for regularity accounts of laws of nature is that they fail to distinguish between genuine laws of nature and merely accidental generalizations. For example, on a necessitarian account, it is a physically necessary truth "that all solid spheres of enriched uranium have a diameter of less than one mile" because it would be physically impossible, given the physical properties of enriched uranium, for such a sphere to exist (Weinert 1995, p. 18). This is neither a logically necessary truth nor an accidental truth. It is true in virtue of the laws of nature. By contrast, "that all solid spheres of gold have a diameter of less than one mile" is an accidental generalization--it may be true throughout the entire history of the universe, but the laws of nature do not forbid the existence of such a sphere (Weinert 1995, p. 18).

Consider another example: Suppose in the history of the universe no Tyrannosaurus rex is longer than 60 feet. According to the necessitarian, this is a purely contingent fact--it just happens to be the case that no T. rex exceeded 60 feet, but it wouldn't be physically impossible for a T. rex to have been a foot longer. By contrast, it would be physically impossible for a photon to have mass. The difference between the two cases is that the laws of nature do not forbid the existence of a T. rex exceeding 60 feet. On a regularity account, however, the laws of nature simply describe truths which hold true throughout the history of the universe. If no T. Rex exceeds 60 feet in the entire history of the universe, the regularity theorist seems forced to describe this truth as a law of nature. On a regularity account, there simply is no distinction between laws of nature and accidental generalizations.

How is all of this relevant to our final criterion for the natural? On a necessitarian account, natural laws describe what must be the case--they limit what is or is not possible. It would be more accurate (and more poignant for our purposes) to say that laws of nature limit what is possible assuming only natural causes are present. So if something occurs which is not possible according to laws of nature, then we have a genuine instance of a supernatural event--a physically impossible event. On a regularity account, however, there can be no violations of natural laws by definition because natural laws simply describe what happens in the whole history of the universe without invoking physical necessity or physical possibility.

If one rejects a necessitarian account of laws of nature in favor of a regularity theory, then the 'behaves in accordance with laws of nature' criterion for the natural will have to be abandoned--on a regularity theory every event that occurs in the history of the universe behaves in accordance with the laws of nature by definition. There simply cannot be a violation of a law of nature on a regularity account. On such an account every event which happens within nature obeys the laws of nature and thus, barring causal overdetermination, has natural causes. On a regularity account, the natural-supernatural distinction is abandoned. I will consider reasons why we should not abandon the natural-supernatural distinction in the next section.

Some laws of nature, such as statistical laws, are merely dispositions or propensities for natural objects to behave in certain ways under certain conditions. Such laws leave room for deviations from the normal course of behavior exhibited by the objects they govern. Nevertheless, the circumstances under which such deviations can occur is still constrained by statistical laws. For example, in quantum mechanics the spontaneous appearance of virtual particles out of a vacuum appears to violate the law of conservation of energy. However, the Heisenberg uncertainty principle allows for the creation of short-lived virtual particles from a vacuum on subatomic scales smaller than the radius of a proton. The creation of a lasting macroscopic physical object such as a chair out of nothing would constitute a violation of a law of nature.

Mental states also appear to be governed by laws of nature. If reductive physicalism is true, mental states obey the same laws of physics that any physical system must obey even if the laws of physics do not make explicit reference to any mental states. If either property dualism or interactionist substance dualism is true, there are laws of nature which govern correlations between mental states and brain states. These laws would not be laws of physics, but psychophysical laws governing, for example, how willing your arm to go up causes it to do so (if mental-to-physical causation occurs) or how being in a certain brain state causes you to feel pain (if physical-to-mental causation occurs).

One objection to identifying the natural with that which behaves in accordance with laws of nature is that the laws themselves, some of which are amenable to scientific investigation, do not behave in accordance with laws of nature. There are two independent issues here. First, scientific laws (our approximations to the laws of nature) are not strictly amenable to scientific investigation in the sense of being scientifically explicable. We cannot provide a scientific explanation for why matter or energy cannot be created or destroyed. We can only explain natural phenomena in terms of such laws. These laws are brute facts where all explanation must ultimately terminate. In the case of our most general laws of nature, we can use science to determine approximations to what these laws are, but not why they obtain rather than a different set of laws. Thus, like qualia appear to be, the laws of nature themselves are in principle scientifically inexplicable. This would provide even stronger grounds for rejecting the 'amenable in principle to scientific investigation' criterion for the natural if we wanted to countenance the laws of nature themselves as part of the natural realm. The second point raised by the objection is that the laws of nature themselves do not obey laws of nature. But this makes our present criterion self-refuting only if the laws of nature are countenanced as part of the natural world. Perhaps they belong to the nonnatural realm. Although laws of nature may necessitate the behavior of natural objects, since by definition they are unchanging they are not causes of change in the behavior of natural objects. Since natural laws cannot be located in space and time and are not causes of change, they appear to qualify as nonnatural abstract objects.

Our central goal in introducing the idea that the natural is that which behaves in accordance with laws of nature was to resolve a difficulty with the 'amenable in principle to scientific investigation' criterion for the natural. That difficulty was that it appears impossible to say a priori whether some phenomenon is amenable to scientific investigation. Only an a posteriori answer is available: In order to determine whether a phenomenon is amenable to scientific investigation we have to subject it to a scientific investigation. If we arrive at a successful scientific explanation of the phenomenon, we have determined that it is amenable to scientific investigation. But for phenomena which continue to resist scientific explanation, we cannot say categorically that they cannot be explained in principle because we have no way of knowing that. There is simply too much room for mistaking the limitations of our cognitive resources or our empirical access to a phenomenon for a metaphysical truth.

The idea that the natural is that which obeys natural laws can explain how something can be amenable to scientific investigation in principle in terms of a more basic property; but it does so at the expense of raising further difficulties associated with the concept of natural laws and does not shed light on the meaning of 'natural' in practice. If the regularity theorist is right, we cannot appeal to the concept of physical necessity to explain events in the natural world. Moreover, even if we assume a necessitarian account of laws of nature, we cannot (from our limited point of view) declare what those laws are with full confidence. Our scientific laws, after all, are merely approximations of genuine laws of nature. They are subject to qualification under different circumstances and we can never exhaust the nearly limitless circumstances under which our scientific laws may need to be qualified. Since no two events are entirely identical, what appears to be a violation of a law of nature may turn out to be, upon further investigation, a new circumstance where our scientific laws need to be modified.

How can we have knowledge of the genuine laws of nature, as opposed to approximations to them? Essentially we can only have knowledge of scientific laws, not the laws of nature themselves, through the empirical discovery of regularities in nature. But if we cannot have knowledge of the laws of nature themselves we cannot say with full confidence that any event is truly a violation of a law of nature (as opposed to a violation of a scientific law). We seem to have reached an impasse in our analysis of the meaning of the term 'natural'. None of our criteria can tell us how to draw the natural-supernatural distinction in practice.

The Natural-Supernatural Distinction

Where does this survey of criterion for the natural leave us? We have seen that there are good reasons for rejecting the simple identification of the natural with the physical, with the spatiotemporal, or with that which is amenable in principle to scientific investigation. In theory, the natural can be identified with the physical or supervenient upon the physical and with that which behaves in accordance with laws of nature. These two criteria can be combined as jointly necessary and sufficient theoretical conditions for the natural. Anything appropriately labeled as natural would have to meet both conditions. Being strictly physical, spatiotemporal, or amenable in principle to scientific investigation would be sufficient conditions (but not necessary ones) for something to be natural. In other words, anything that was physical, spatiotemporal, or scientifically explicable would clearly be natural, but these are not requirements something must meet in order to be considered natural. Conversely, however, being nonphysical, nonspatiotemporal, and scientifically inexplicable would be necessary but not sufficient conditions for the nonnatural--anything that is physical, spatiotemporal, or scientifically explicable cannot be nonnatural.

Combining these criteria with a negation of our remaining criteria for the natural, we are left with the following jointly necessary and sufficient theoretical conditions for the nonnatural: (1) it is nonphysical and not supervenient upon the physical; (2) it is nonspatiotemporal; (3) it is scientifically inexplicable in principle; and (4) it fails to behave in accordance with natural laws. Abstract objects, as paradigm cases of nonnatural objects, meet all of these conditions. Since the supernatural is a more specific subclass of the nonnatural restricted to nonnatural causes of events within nature, a supernatural event would have to be an event in the natural world which has a cause meeting all four of the conditions given above. In theory, causes of events in the natural world which meet these criteria are supernatural causes. But, as we have seen, it is impossible to apply any of these theoretical criteria in practice because of our limited epistemological resources in answering metaphysical questions. We simply do not know what sorts of objects, properties, or phenomena count as nonphysical, what sorts resist scientific investigation in principle, or which actually behave in accordance with genuine laws of nature. Does this mean that we should abandon the natural-supernatural distinction?

Given that we can formulate self-consistent jointly necessary and sufficient theoretical conditions for both the natural and supernatural, the natural-supernatural distinction can be coherently spelled out in theory. However, this does not mean that the distinction has any validity in practice. Labeling an occurrence a supernatural event implies that the "the power and laws of nature could not bring [that event] about" (Hepburn 1972, p. 454). However, the objection goes, we are not in the epistemological position to make such a determination. As Ronald Hepburn points out,

It is anything but easy... to elaborate coherently the nature-supernature distinction. Crucial to it is the claim that we can distinguish what lies within the capacities of nature from what lies beyond them. Our knowledge of nature's powers and laws is itself derived from our experience and observation of events. What we judge to be possible depends upon what we have reason to believe actually occurs or has occurred... [W]hat shall we do with the happenings that, eventually, we wish to label miraculous? To exclude them would be to imply that we already know what nature's powers are, that there are criteria prior to experience by which we interpret our observations. But to include them makes it impossible for us to treat them later as miraculous exceptions to natural laws (Hepburn 1972, p. 454).

Hepburn points out a crucial dilemma: Either we retain the natural-supernatural distinction in practice, placing ourselves in the difficult position of trying to determine a priori which sorts of phenomena are physically impossible--something we cannot say with full confidence--or we abandon the practical distinction altogether, thereby making 'naturally-caused event', for all practical purposes, equivalent to 'any event which occurs in the natural world'. Given these two forced options, we should opt for retaining the natural-supernatural distinction in practice simply because the cost of abandoning it is too great.

Naturalism would become trivially true if we abandoned the practical natural-supernatural distinction. The argument for trivial naturalism would be that it is impossible to draw the natural-supernatural distinction in practice and therefore we are never justified in positing supernatural causes because we can never say that the cause of an event is not natural. This implies that naturalism always wins by default on epistemological grounds, even though not necessarily so on metaphysical grounds. Naturalism becomes trivially true because no matter what phenomenon we encounter, we cannot say it is not a naturally-caused phenomenon. If we abandon the natural-supernatural distinction in practice, so far as we can tell any logically possible event may be a naturally-caused event. Consequently, naturalism becomes trivially true in practice but uninformative and supernaturalism is never an option for belief since all causes of events in the natural world may be natural causes.

Hepburn's point--that we simply do not know what is within nature's powers or laws--is borne out in every impasse we have hit in trying to determine how to apply our theoretical criteria for the natural and the supernatural in practice. While we have a pretty good idea of what 'natural' means in theory, it is difficult to say in practice what counts as instance of an event with a natural or supernatural cause. Herbert Spiegelberg argues that because the supernatural (as a species of the nonnatural) is defined in wholly negative terms, it has little--if any--positive content. If by a supernatural event we mean not merely an extraordinary event but an event that is beyond the power of nature to produce, requiring a suspension or violation of the laws of nature, Spiegelberg points out:

Now it takes very little to see that even this conception of a supernatural event is predominantly negative. It implicitly denies that things usually happen this way and that we, with our knowledge of nature, can ever hope to comprehend them. But we learn nothing positive about the nature of the miraculous event... [W]e are faced with a merely negative understanding of the supernatural (Spiegelberg 1951, p. 348).

But what does a negative understanding of a concept imply? Spiegelberg notes that negative concepts refer to the absence of specific characteristics. For example, the term 'nonwhite' refers to all colored things except those which are white. While 'nonwhite' is a negative concept, it indirectly refers to the positive denotative complement of the term white--colored things. Because 'white' is a species of the proximate genus of colored things, nonwhite too is a species of colored things. The problem Spiegelberg sees with the concept of the supernatural is that there is no proximate genus to which both the natural and nonnatural refer:

[I]n an example like 'not-white'... after subtracting white, all the other known colors remain as, what one might call, the denotative complement. But what exactly is the denotative complement in the case of the not-natural, once we have subtracted the whole range of natural objects and events? It is at this point that the heart of the difficulty appears: we cannot point out a denotative complement to the name 'natural' (Spiegelberg 1951, p. 354).

Spiegelberg argues that because we literally know nothing about the nonnatural we cannot find any positive attributes common to both the natural and nonnatural (Spiegelberg 1951, p. 354). While denying that supernatural terms are completely meaningless, Spiegelberg concludes that "they have no more meaning than an expression like 'the 100th dimension' or '100-valued logic'" (Spiegelberg 1951, p. 355).

Despite this conclusion, Spiegelberg concedes that supernatural terms may have what he calls diagnostic meaning. Diagnostic meaning refers to identifying criteria "for recognizing cases of the supernatural when we see them and for deciding whether a specific pretender to the supernatural title has any right to it" (Spiegelberg 1951, p. 355). However, Spiegelberg is skeptical that the wholly negative concept of the supernatural is meaningful enough to provide means for the reliable identification of the supernatural:

As long as we are unable to point out as much as one positive complement to all the natural properties excluded by the supernaturalist we would never have a chance to tell with assurance: 'This event must be a case of the supernatural... an intercession from a world outside nature into nature' (Spiegelberg 1951, p. 356).

I agree that once we subtract out the natural, there is no positively characterized ontological category to which the nonnatural or supernatural belongs. This does make it impossible to positively confirm that an event had a supernatural cause in the way that we can confirm that engine failure caused a plane to crash. However, this does not strip the supernatural of diagnostic meaning altogether. All that is needed for diagnostic meaning for the supernatural is some reasonable criteria for what a supernatural event would look like. We do not need to say that something meeting those criteria must be supernatural; we need to say only that a supernatural event must meet those criteria. This does, of course, leave open the possibility that something entirely natural could also meet those criteria. But, as we shall see, this concession is not necessarily fatal to drawing out the natural-supernatural distinction in practice. It allows us to distinguish a class of clearly naturally-caused events from a class of possibly supernatural events. While this does not provide a definitive classification for the supernatural, it does narrow down our possible candidates for a supernatural event.

Spiegelberg gives two reasons for denying that we can secure indirect evidence for the supernatural in the way that we can secure indirect evidence for electrons. First, we can only devise testable predictions from well-defined positive concepts (Spiegelberg 1951, p. 356). Second, it is impossible to confirm conclusively that any event has a supernatural cause: "The natural signs predicted as indications of the supernatural are always so indefinite that they can never constitute conclusive proof either way. No confirmation is possible, since no disconfirmation can be conceived of that would unmistakably settle the case" (Spiegelberg 1951, p. 356).

Consider Spiegelberg's second reason for denying that we can secure indirect evidence for the supernatural. I agree that we cannot provide absolutely certain evidence for the supernatural. However, I contend that, in principle, we could provide evidence that any reasonable person would regard as conclusive evidence of the supernatural. First, we can provide conclusive evidence that a certain event occurred. The evidence that an event occurred will clearly be of a different sort than experimental evidence for a scientific hypothesis. The evidence for a scientific hypothesis often consists of repeatable results obtained under carefully controlled conditions, whereas the evidence that any sort of past event occurred is not going to be repeatable on demand in this way. Nevertheless, the evidence that an event occurred can be strong enough that it would be irrational to deny that the event occurred. For example, we have very reliable evidence that a nuclear bomb fell on Hiroshima in 1945, evidence that goes beyond even simple historical records, such as eyewitness testimony, video footage, radioactive traces, etc.

Second, we can provide very good reasons for denying that a certain kind of event could plausibly be construed as an event with a natural cause. Some imaginable events, if they occurred, would be so suggestive of a supernatural cause that it would be unreasonable to hold out indefinitely for a natural explanation of them. I will defend this point in more detail when I present my criteria for drawing out the natural-supernatural distinction in practice. Contrary to Spiegelberg's assertion, there is no reason in principle why there couldn't be conclusive evidence for an event extremely suggestive of a supernatural cause. Moreover, even if we could say only that an event was suggestive of a supernatural cause, though not conclusively so, Spiegelberg is wrong to assert that disconfirmations are not possible. A successful natural explanation of an allegedly supernatural event would be as much disconfirmation of a supernatural cause as the evidence for evolution is disconfirmation of six-day creationism.

Spiegelberg's first reason for denying that we can secure indirect evidence for the supernatural--that we cannot derive testable predictions from supernatural explanations--is related to the following objection: "By what method, if any, can we expect to identify an instance of [supernatural causation] in reality? Without such a method we would be cut off from any possibility of genuine knowledge" (Spiegelberg 1951, p. 357). Spiegelberg's point is that, unlike successful naturalistic explanations grounded in scientific method, we have no methodology on which to ground supernatural explanations.

While I agree with Spiegelberg that, unlike natural explanations, we have no grounds for evaluating competing supernatural explanations for an event, this does not mean that we cannot identify a likely candidate for a supernatural event. I contend that, in principle, there cannot be any methodology which would provide reliable supernatural explanations of an event because of the nature of supernatural causation. Supernatural causation requires a supernatural cause of a natural effect. But, by definition, we cannot have empirical access to a supernatural cause. We can have empirical access only to the effect of a supernatural cause. In order to provide reliable explanations of phenomena we must establish causal relationships through scientific investigation. But since we have no empirical means for gaining information about supernatural causes themselves, we cannot establish laws which express causal relationships between supernatural causes and natural effects, should there be any. We need access to both cause and effect to establish any sort of reliable causal explanation. However, this means only that we cannot understand how a supernatural event occurred; it does not mean that we cannot say that it did occur.

Spiegelberg maintains that ordinarily we establish the existence of an instance of a negative concept by establishing the existence of part of its positive denotative complement--e.g. to establish the existence of 'not-white' we establish the existence of some other color. But since we can find no positive denotative complement to the nonnatural or supernatural, it will be extremely difficult to establish an instance of it:

Since we found that the supernatural in its prevalent negative interpretation has no such positive denotational complement, this simple method of verification is out of the question. Suppose, then, some object should present itself as a candidate to supernatural dignity... All we can hope to do, if we want to check such a claim critically, is to show that it does not fit into any known category of nature. We have therefore to make sure of a comprehensive review of the natural realm, running through all its known and conceivable categories with a view to establishing at least indirectly that here is unmistakable evidence of something different from the familiar pattern and not covered by our 'natural' categories. To do this exhaustively may not seem impossible on principle. But on any concrete occasion this is certainly no small assignment (Spiegelberg 1951, p. 358).

Although establishing an instance of supernatural causation may not be impossible in principle, Spiegelberg contends that to do so in any particular case one would have to eliminate all possible natural explanations for an event--including scientific explanations cast in terms of a future science not yet available to us. Of course, this is not a practical possibility. But even if we could somehow pull this off, Spiegelberg further argues that appeals to the supernatural in such cases, given the negative character of the supernatural, simply reduce to appeals to ignorance.

Given the lack of any methodology for evaluating competing supernatural explanations and the lack of any empirically compelling claim to revelation, I agree that the most we could say in such cases is that the cause of the event in question is not natural. We couldn't say anything about the nature of the supernatural cause itself. In this sense any particular supernatural explanation would be an appeal to ignorance. However, when no natural cause for an event can be found after a completely exhaustive search for natural causes, our only apparent recourse is to postulate a supernatural cause for the event. In actual cases of a possible supernatural event, of course, we will not be able to conduct a completely exhaustive search for natural causes. We cannot do this even for events with natural causes. Other considerations I will consider briefly will make it clear that Spiegelberg's requirement of an exhaustive search of the natural world to establish that a supernatural cause of an event is far more likely than a natural one is too strong.

Spiegelberg's fundamental claim is that in order to say that a phenomenon is supernatural with assurance we must have an exhaustive knowledge of the natural realm and what is possible within it:

There is, however, a much more presumptuous claim involved in any serious affirmation of the supernatural. The supernatural interpretation presupposes a knowledge of the natural sphere so clear and so exhaustive that we can be absolutely sure it does not offer any room for the phenomenon under consideration... Whoever asserts that a certain event is supernatural also asserts that he has complete knowledge of the field marked off as nature as well as full knowledge of what is and what is not possible within it (Spiegelberg 1951, p. 358).

Do we actually have to have complete knowledge of the natural world in order to distinguish between an event with a natural cause and an event with a supernatural one? If we want complete assurance that an event has a supernatural cause we may need an exhaustive survey of the natural world. However, Spiegelberg does not address the possibility that we could determine that an event has a supernatural cause with a reasonable degree of confidence. I will now defend the position that we are in a position to identify events with supernatural causes with a reasonable degree of confidence.

It is difficult, but not impossible, to draw the natural-supernatural distinction in practice. Our impasse in determining what counts as an event with a supernatural cause does not reflect a deficiency in our theoretical criteria for the supernatural, but rather the practical limitations of our epistemological access to fundamental metaphysical categories like 'physical', 'laws of nature', etc. We simply cannot say categorically what it means for something to be physical, to be amenable in principle to scientific investigation, or to be a law of nature. This is a practical difficulty that can be overcome by replacing our theoretical criteria for the supernatural with our closest approximations to them in practice. Thus, what we need is a less theoretical set of conditions for events which may have supernatural causes--what I will call criteria for a likely candidate for a supernatural event. Given that we are working with approximations to our theoretical criteria, it is possible, though unlikely, that something considered entirely natural in terms of our theoretical criteria would be consistent with our practical criteria for a likely candidate for a supernatural event. In other words, anything meeting these practical criteria could be entirely natural, but it is unlikely that anything meeting all of them would be. What would these in practice criteria look like?

In order to answer this question, we must ask another: What would be an indication that an event has a supernatural cause? Consider three examples of what many people would regard as clear-cut supernatural events--stars lining up to spell out 'God exists', the resurrection of a corpse after a century of decomposition, and a marble statue waving its hand at you. What is it about these examples that makes us think these events cannot have solely natural causes? Is it simply that they do not ordinarily happen, much like extinction-level meteor impacts with the Earth? Or is it something more?

In addition to meeting approximations of our theoretical criteria for the supernatural, these examples seem to display other commonalities. Our theoretical criteria for the nonnatural consisted of the following jointly necessary and sufficient conditions--the nonnatural must be nonphysical and not supervenient upon the physical, be nonspatiotemporal, be scientifically inexplicable in principle, and fail to behave in accordance with natural laws. Our practical approximations to these theoretical criteria combined with the commonalties in the 'clear-cut' examples provide the following jointly necessary and sufficient conditions for a likely candidate for a supernatural event: (1) the cause of the event cannot be identified as any known physical force or entity nor is it supervenient upon any known physical force or entity; (2) the cause of the event cannot be located in space and time; (3) the event defies all attempted scientific explanations thus far; (4) the event appears to violate well-established scientific laws (as distinguished from genuine laws of nature); (5) the event is highly improbable if it solely has known natural causes; and (6) the event exhibits apparently purposive or intelligent behavior. Any event meeting these criteria is a likely candidate for a supernatural event. Note that being a likely candidate for a supernatural event is a necessary but not a sufficient condition for actually being a supernatural event.

The first three conditions for a likely candidate for a supernatural event seem relatively straightforward. The fourth and fifth conditions are more complicated and related in fundamental way. Note that it would be a stronger approximation to our theoretical criteria to say that a likely candidate for the supernatural violates scientific laws rather than appears to violate scientific laws. However, even given known scientific laws, it isn't always clear what constitutes a violation of a scientific law. In theory, an event with a supernatural cause really would be physically impossible--given fixed initial conditions, genuine laws of nature, and only natural causal factors, a physical system would have to behave in a certain way. This is true even given physical indeterminism--some outcomes will be more probable than others, but some outcomes will be excluded altogether. For example, while the Heisenberg uncertainty principle allows the creation of short-lived subatomic particles out of nothing, the creation of a macroscopic physical object out of nothing would be a violation of a law of nature. A physically impossible event would be an event which behaved contrary to what would be predicted given initial conditions, genuine natural laws, and all relevant natural causal factors. Because we often cannot know the initial conditions and relevant natural causal factors involved in bringing about an event even assuming that scientific laws are genuine laws of nature, a physically impossible event may not involve a straightforward violation of a scientific law.

For example, a resurrection can be described as a localized reversal of entropy, something which is not explicitly ruled out by scientific laws. However, given likely initial conditions (e.g. an advanced state of decomposition) and all likely relevant natural factors (e.g. the destruction of most of the physical information about the body as it was before death due to decomposition and the lack of an advanced replication machine), some events may be so improbable that the chances of such events occurring even once given our background knowledge of the relevant natural causal factors produces astronomical odds. Consider Richard Dawkins' example of a statue waving its hand at you:

In the case of the marble statue, molecules in solid marble are continuously jostling against one another in random directions. The jostlings of the different molecules cancel one another out, so the whole hand of the statue stays still. But if, by sheer coincidence, all the molecules just happened to move in the same direction at the same moment, the hand would move. If they then all reversed direction at the same moment the hand would move back. In this way it is possible for a marble statue to wave at us. It could happen. The odds against such a coincidence are unimaginably great... The number is so large that the entire age of the universe so far is too short a time to write out all the noughts! It is theoretically possible for a cow to jump over the moon with something like the same improbability (Dawkins 1987, pp. 159-160).

Dawkins argues that if such an improbable event occurred we should regard it as a supernatural event "because all our experience and knowledge tells us that marble doesn't behave like that" (Dawkins 1987, p. 159). These events do not straightforwardly involve violations of scientific laws because, in practice, all relevant natural causal factors are not given. If they were all known, however, then a resurrection would almost certainly involve a violation of scientific laws. Because of the limitations of our knowledge, we will have to work with likely initial conditions and likely natural causal factors. Certain sets of causal factors will be more likely to obtain than others. But because we do not know that a scientific law necessarily is violated (e.g. a levitation could result from some other natural force counteracting the law of gravity, such as magnetic repulsion), for all practical purposes, an event appearing to violate scientific laws reduces to a highly improbable event given only known kinds of natural causes. Note that an event which is highly improbable given known natural causes is not necessarily an infrequent event. If supernatural causes are present, we would expect a greater frequency of events which would be very unlikely if only natural causes were present. In order for this distinction to hold, we must assume something like a propensity interpretation of probability instead of a relative frequency interpretation--that is, we must assume that the probability of an event refers to a propensity of a physical system to produce a given outcome.

The last condition for a likely candidate for a supernatural event--that the event exhibits apparently purposive or intelligent behavior--is fairly ambiguous. Note that this final condition is the only condition which cannot ultimately be derived from our theoretical criteria for the nonnatural. Purposive or intelligent behavior contrasts with the way natural phenomena are often characterized--as impersonal or the result of impersonal, unintelligent causes. Natural forces are typically characterized this way because the basic natural causes which constitute the subject matter of physics are impersonal and unintelligent and physics has given us the most reliable conclusions of all the sciences. Of course this condition cannot be derived from our theoretical conditions for the natural or supernatural precisely because some natural objects like animals or even heat-seeking missiles exhibit purposeful behavior. Purposeful or intelligent behavior is not necessarily a sign of supernatural causation because natural objects also exhibit it. But our consideration of the three examples of 'clear-cut' supernatural events all included this element and I think we can tentatively conclude that no supernatural event would have impersonal causes and thus purposeful behavior should be one of the conditions for a likely candidate for a supernatural event.

We can elaborate on what is required for an event to meet the final condition by providing examples of events which do meet it. However, to back up the claim that it is unlikely for an event with a natural cause to meet all of the criteria I have outlined for a likely candidate for a supernatural event, I will consider an example of an event meeting all six conditions. Imagine that a religious leader claimed that he would return to vindicate his followers' faith that he is the savior of the world a half century after his death. Imagine that this leader made arrangements with his followers that his corpse would be kept on display after his death. Anticipating his resurrection and with vast financial resources, the followers ensured that there would be a substantial media presence at the displayed tomb fifty years later to the day. While the journalists reluctantly appeared at the site, fulfilling a contract, none of them seriously believed that this resurrection would occur. The leader's followers also ensured that a DNA test would be performed on the spot that day to confirm the identity of the corpse.

An hour after the DNA test results confirmed the corpse's identity, and while television cameras rolled, the corpse spontaneously transformed into a likeness of the original man. The likeness of the leader then turned to the astonished reporters and cameramen and said "I will bring peace and harmony to the world and heal the sick". At the moment those words were spoken, every human being on the planet lost all symptoms of disease. The likeness of the leader then said "I will return in 6 months" and disappeared. An extensive worldwide UN-sponsored medical check-up confirmed that no examined individual on the planet had any trace of disease.

This would clearly be a likely candidate for a supernatural event. The cause of the transformation and universal healing could not be identified as any known physical or spatiotemporal cause, would be scientifically inexplicable, would appear to violate natural laws, would be highly improbable given only known kinds of natural causes, and the transformed corpse's communication and apparent healing powers would exhibit purposeful behavior. I contend that in the hypothetical example just given the well-documented events could not plausibly be regarded as events with solely natural causes.

Throughout human history, supernatural causes have been invoked to explain droughts, earthquakes, thunderstorms, comets, the spread of disease, mental illnesses, mystical experiences, the orbits of the planets, the origin of living things, and the origin of the world, among many other phenomena. As the scientific revolution of the 17th and 18th centuries flourished, appeals to supernatural causation ultimately gave way to successful scientific explanations of various phenomena in terms of natural causes. Ever since its inception, science has increasingly strengthened the plausibility of naturalism by providing informative accounts of a wide range of phenomena in terms of natural causes. The more science has progressed, the less room there has been for postulating supernatural causes within a scientific account of the world, and if past experience is any guide, the trend will continue well into the future. This trend has led many to conclude that there probably are no genuine instances of supernatural causation. As science explains more of the natural world around us, appeals to supernatural causation become less plausible.

Many philosophers and scientists have concluded that the best explanation for our ability to develop successful scientific explanations for such a wide range of phenomena in terms of natural causes is that there are no genuine instances of supernatural causation. Barbara Forrest, for example, describes naturalism as "a generalization of the cumulative results of scientific inquiry" (Forrest 2000, p. 19). In other words, the best explanation for the success of science is that naturalism is true. Given the proliferation of successful scientific explanations for phenomena, Forrest concludes that there is "an asymptotic decrease in the existential possibility of the supernatural to the point at which it is wholly negligible" (Forrest 2000, p. 25). If naturalism were false, there would be some phenomena that could not be explained solely in terms of natural causes. However, because science can explain all of the uncontroversial phenomena we have encountered in terms of natural causes, there probably are no phenomena which cannot be explained in terms of natural causes. Therefore, naturalism is probably true.

This success of science argument rests on a crucial inductive premise--that we can infer that all phenomena can be explained in terms of natural causes from the ability of science to explain all of the uncontroversial phenomena we have encountered in terms of natural causes. Even if we accept the validity of this inductive inference, we still have to establish that all the uncontroversial phenomena we have encountered so far can be explained scientifically. Since there certainly are uncontroversial phenomena for which we lack successful scientific explanations--consider the prevalent gravitational influence of some unknown form of dark matter in the universe--I will defend a related but stronger argument for naturalism. This argument does not require us to have a successful scientific explanation for all well-established events in order to provide evidential support for naturalism.

A likely candidate for a supernatural event is not necessarily the result of supernatural causation given that meeting the criteria for a likely candidate is a necessary but not a sufficient condition for actually being a supernatural event. Thus, if naturalism is true, it does not necessarily follow that there will be no likely candidates for a supernatural event--it is possible, however unlikely, that a naturally-caused event would also meet the requirements for a likely candidate for a supernatural event. For example, suppose that a subject can induce out-of-body experiences at will in a laboratory setting. During several experimental trials, after this subject has induced an out-of-body experience, infrared cameras capture the outline of a person moving toward a bell which begins to ring in a room adjacent to the location of the subject's normal physical body. If such events occurred today, they would meet all of the criteria for a likely candidate for a supernatural event. Nevertheless, such events might be the result of entirely natural causes which could be understood only in terms of some future science not yet available to us. For example, one might postulate that human organisms possess natural astral bodies made of some unknown form of exotic matter which can detach from normal physical bodies in certain circumstances. In the absence of successful scientific explanations for such phenomena, however, uncontroversial instances of likely candidates for a supernatural event would make supernaturalism more likely to be true than not relative to a background scientific picture lacking natural categories for such events.

Regardless of such possibilities, if there are any events within nature that have supernatural causes, these events will be likely candidates for a supernatural event. Thus, if naturalism is false, there will be events which are likely candidates for a supernatural event. Even without a definitive set of criteria for identifying a supernatural event, we can see the beginnings of an argument for naturalism:

(P1) If naturalism is false then there are events which are likely candidates for a supernatural event.