Methodological naturalism

Methodological naturalism is the procedural principle that guides scientific inquiry: when scientists seek to explain natural phenomena, they restrict themselves to natural causes, natural processes, and natural entities. The principle does not assert that nothing supernatural exists — that would be the separate and much stronger claim of philosophical naturalism or metaphysical naturalism. It asserts only that science, as a practice, is not equipped to detect, test, or evaluate supernatural explanations, and that invoking them terminates inquiry rather than advancing it.^{4, 13} Understanding this distinction — between a methodological commitment and a metaphysical one — is essential to evaluating both the scope and the limits of scientific knowledge, and to assessing the recurring charge from critics that science is inherently biased against the possibility of God.

The principle is so deeply embedded in modern scientific practice that it often goes unstated, yet it has a traceable intellectual history and has been the subject of sustained philosophical analysis. Its defenders argue that it is not an arbitrary exclusion but a pragmatic requirement: any explanation that appeals to an agent capable of acting without constraint, without pattern, and without predictability is not an explanation at all in the scientific sense — it is a suspension of explanation.^{5, 18} Its critics, most notably Alvin Plantinga, argue that the commitment to naturalism in science smuggles in philosophical assumptions that are not themselves scientifically warranted. This debate touches the deepest questions in the philosophy of science, the relationship between science and religion, and the logic of god-of-the-gaps reasoning.

The MN–PN distinction

The most common misunderstanding of methodological naturalism is the equation of it with philosophical naturalism. Philosophical naturalism — sometimes called metaphysical naturalism or ontological naturalism — is the view that nothing beyond the natural world exists: there are no gods, no souls, no supernatural entities or forces of any kind.¹⁴ This is a substantive metaphysical position, one that entails atheism and is itself beyond the reach of scientific proof or disproof. Methodological naturalism, by contrast, makes no such claim. It is compatible with theism, agnosticism, deism, and every position that does not require science itself to confirm the supernatural.

The philosopher Robert Pennock, who has written extensively on this distinction, frames it precisely: a scientist who practises methodological naturalism is not asserting that God does not exist; she is asserting only that her laboratory, her statistical tools, and her peer-reviewed methods are not adequate instruments for detecting God’s actions.^{4, 5} This is not a statement about God’s ontological status; it is a statement about the limits of a particular set of tools. The historian of science Ian Barbour makes a related point: the most productive approach to science and religion recognises their different but complementary magisteria, and resists the mistake of treating a methodological commitment of science as a verdict on religion.¹⁹

The practical upshot is significant. Scientists who are devout theists — a substantial proportion of the scientific community across all historical periods — have no difficulty operating under methodological naturalism because they do not interpret it as a denial of their faith.⁸ Francis Collins, Kenneth Miller, Mary-Claire King, and countless others have maintained orthodox religious belief while conducting science entirely within the naturalistic framework. The MN–PN distinction is what makes this coherent: one can believe that God created the natural order while also believing that the natural order is, for scientific purposes, self-sufficient and amenable to purely naturalistic investigation.

Historical development

The explicit articulation of methodological naturalism is a product of early modern science, but its roots lie in the broader intellectual movement that sought to disentangle natural philosophy from revealed theology. Francis Bacon’s Novum Organum (1620) laid the groundwork by insisting that knowledge of nature must be derived from observation and inductive reasoning, not from received authority, scripture, or metaphysical speculation.¹ Bacon was himself a believing Christian, and he explicitly distinguished between the book of nature and the book of scripture, arguing that conflating them did a disservice to both. His method was not anti-religious; it was a discipline of attention, an insistence that natural facts be established by natural investigation.

Isaac Newton represents a more complex case. His Principia Mathematica (1687) achieved the most spectacular early triumph of the naturalistic method, deriving the motions of the heavens from a small set of mathematical laws. Yet Newton himself was not a pure methodological naturalist: in the General Scholium added to the second edition (1713), he invoked God to explain the stability of the solar system, arguing that the mutual gravitational perturbations of the planets would eventually disrupt their orbits and that God must periodically intervene to correct them.² His famous phrase “hypotheses non fingo” — I frame no hypotheses — expressed a commitment to inferring causes only from observed phenomena, a principle consistent with MN, yet Newton himself crossed the line he described when naturalistic tools ran out.

A century later, Pierre-Simon Laplace closed Newton’s gap. Using perturbation theory, Laplace demonstrated in his Mécanique Céleste (1799–1825) that the solar system is self-stabilising over long timescales without any external correction.³ The anecdote in which Napoleon reportedly asked Laplace where God fit into his system, and Laplace replied “I had no need of that hypothesis,” became the paradigmatic statement of MN as a working scientific attitude — not an atheist manifesto but a scientist’s declaration that his explanatory machinery was complete without divine intervention. The episode also marks the birth of the god-of-the-gaps pattern: Newton had filled a genuine gap in his account with divine action; Laplace showed the gap was an artefact of incomplete mathematics.

Charles Darwin’s On the Origin of Species (1859) extended the naturalistic program decisively into biology.¹⁷ Before Darwin, the appearance of design in living organisms was the strongest apparent counterexample to methodological naturalism: no purely natural process seemed capable of generating the intricacy of an eye, a wing, or a hand. Darwin’s mechanism of natural selection acting on heritable variation demonstrated that cumulative adaptation required neither foresight nor a designing intelligence. This did not prove that God had not designed life; it demonstrated that a naturalistic account was available and testable. The explanatory work previously done by appeals to divine design was transferred to a mechanism that could be observed, modelled, and falsified.²⁰

Why MN is necessary for science

The necessity of methodological naturalism for scientific practice follows from the basic requirements of testability and falsifiability that define scientific explanations. An explanation is scientifically useful to the degree that it generates predictions that can be checked against evidence: if the prediction fails, the explanation is falsified or requires revision; if it succeeds, it gains evidential support. A supernatural explanation — one that invokes an agent unconstrained by natural laws — cannot in principle generate such predictions, because a supernatural agent is by definition free to do anything, and an agent that can do anything predicts everything and therefore predicts nothing.^{13, 18}

This is not a trivial objection. If science were to allow “God did it” as an explanation for any phenomenon, then every unexplained observation could be resolved by that single appeal, and the incentive to search for natural mechanisms would be eliminated. The scientific method derives its power from the fact that it accepts explanations only when they can be tested and, in principle, refuted. An explanation immune to refutation is not a scientific contribution; it is an explanation-stopper.⁵ Maarten Boudry and colleagues have described this as the “god-stopper” problem: supernatural explanations, if admitted, would foreclose exactly the kind of inquiry that has produced the accumulated body of scientific knowledge.¹⁸

There is also a practical argument from success. Every scientific field has a track record of replacing explanations grounded in ignorance or supernatural appeal with explanations grounded in natural mechanisms. Infectious disease, once attributed to divine punishment or demonic influence, was explained by germ theory. Mental illness, once attributed to demonic possession, was progressively explained by neurology, biochemistry, and psychology. Lightning, earthquakes, eclipses, the motion of planets, the diversity of species — all were once attributed to divine agency, and all received naturalistic accounts that were more predictive, more precise, and more practically useful than the accounts they replaced.⁸ The direction of this history is not incidental. As Samir Okasha notes, the progressive success of naturalistic explanation across domains provides a strong inductive warrant for MN as a heuristic, independent of any philosophical argument for its necessity.¹³

Thomas Kuhn’s analysis of scientific paradigms is also relevant here. Scientific communities operate within paradigms that define what counts as a proper explanation, what anomalies are tolerable, and when a theory is overthrown.¹⁵ MN functions as part of the paradigm of modern science: it is not a rule written into a textbook but a shared professional commitment that shapes what scientists regard as satisfying explanations. Departing from it is not simply a logical option; it would require abandoning the explanatory standards that make science recognisably science.

Kitzmiller and the legal status of MN

The most consequential judicial ruling on methodological naturalism came in the 2005 case Kitzmiller v. Dover Area School District, in which Judge John E. Jones III of the United States District Court for the Middle District of Pennsylvania ruled that intelligent design is not science and may not be taught in public school science classes as an alternative to evolution.⁶ The ruling addressed methodological naturalism directly and at length.

Judge Jones found that intelligent design violates MN because it “invokes and permits supernatural causation” and thereby abandons “the ground rules of science.” He identified five criteria that demarcate science from non-science — testability, peer review, error correction, consistency with accepted scientific consensus, and commitment to natural explanations — and found that ID failed each of them.⁶ The opinion cited testimony from philosophers of science, biologists, and historians of science to establish that the commitment to naturalistic explanation is not a bias of individual scientists but a structural feature of the scientific enterprise as it has operated since the seventeenth century. Supernatural explanations had been progressively excluded from science not because scientists were prejudiced against religion, but because such explanations proved scientifically sterile.

The ruling noted that MN does not prevent individual scientists from holding religious beliefs, and it does not constitute the government endorsing atheism. It simply identifies the methodological ground rules without which “science” would cease to be a coherent category. Critics of the ruling, particularly in the intelligent design movement, charged that Jones had confused MN with metaphysical naturalism and had imposed a philosophical assumption on science rather than identifying a feature of it. The philosopher of science Samir Okasha, among others, has responded that this objection misconstrues the argument: the case for MN rests on the practical requirements of testability, not on the metaphysical claim that nothing supernatural exists.¹³

Creationist and ID objections

Critics of methodological naturalism from creationist and intelligent design camps have raised several objections, most of which resolve into two main arguments: that MN constitutes an arbitrary philosophical bias, and that it systematically excludes relevant evidence.

The “arbitrary bias” objection is associated most prominently with Philip Johnson, a law professor at Berkeley who became the intellectual godfather of the intelligent design movement. Johnson argued in works including Defeating Darwinism by Opening Minds (1997) that MN is a philosophical presupposition smuggled into science, one that guarantees a naturalistic conclusion before any evidence is examined.¹⁶ If you assume in advance that only natural explanations will be considered, Johnson contended, then you will always find a natural explanation or, where you cannot, simply declare the problem unsolved rather than entertain a supernatural answer. The deck is stacked.

The response to this objection is that MN is not a presupposition about what the correct answer is; it is a constraint on what kinds of answers are scientifically evaluable.^{4, 5} The analogy would be objecting to the rules of chess on the grounds that they “arbitrarily” exclude moving pieces off the board: the rules do not bias the game toward any particular outcome; they define what the game is. MN defines what science is. Moreover, Johnson’s objection would apply equally to any methodological constraint: one could equally claim that the requirement for peer review is an “arbitrary” bias toward ideas that peers find plausible. The question is whether the constraint serves the goals of the enterprise, and in the case of MN the historical track record provides a strong affirmative answer.

The “excluding evidence” objection holds that MN prevents science from following the evidence wherever it leads, and that if the evidence for design is strong enough, science should be willing to revise its method to accommodate it. Michael Behe, whose concept of “irreducible complexity” forms the core of biochemical ID arguments, has pressed this point: if it can be demonstrated that certain biological systems could not have been assembled by any known natural process, then the inference to a designing intelligence should be scientifically admissible regardless of MN.¹¹

The difficulty with this argument is threefold. First, the inference “no known natural process could produce this” is an argument from ignorance, not a demonstration of impossibility; the history of science is a long record of “no known natural process” claims that were subsequently falsified. Second, even if one grants that a system is not producible by natural processes as currently understood, the inference to an intelligent designer requires a further step that MN correctly flags as scientifically unavailable: there is no way to test or characterise the claimed designer, specify what it would or would not produce, or generate falsifiable predictions from the design hypothesis.¹² Third, the Judge Jones ruling in Kitzmiller documented in detail how the specific examples advanced by ID proponents — the bacterial flagellum, the blood clotting cascade — had received detailed naturalistic accounts in the peer-reviewed literature, deflating the empirical premise of the argument.⁶

Plantinga’s critique

The most philosophically sophisticated challenge to methodological naturalism comes from the Reformed epistemologist Alvin Plantinga, particularly in his 2011 book Where the Conflict Really Lies: Science, Religion, and Naturalism. Plantinga’s argument is not the populist one of Johnson; it is a careful philosophical analysis directed at the internal coherence of MN.⁷

Plantinga distinguishes between what he calls “Duhemian MN” and “Augustinian MN.” Duhemian MN is the view that science should restrict itself to natural explanations for practical methodological reasons — because they are testable, because they advance inquiry, because they are the only kind of explanation the scientific community can evaluate collectively. Plantinga largely accepts this as a reasonable working principle. His objection is directed at a stronger version, which he calls the claim that MN must be adopted as a matter of philosophical necessity, rather than practical convenience. If MN is merely a useful heuristic, it might in principle be relaxed when the evidence is sufficiently compelling. If it is a necessary feature of science, it is doing metaphysical work while pretending not to.⁷

Plantinga also connects his critique to his broader Evolutionary Argument Against Naturalism (EAAN), which is addressed separately in the article on the evolutionary argument against naturalism. The EAAN argues that a fully naturalistic worldview undermines the reliability of the cognitive faculties it depends on, producing a self-defeating position. If this argument succeeds, then philosophical naturalism is incoherent, and the alleged entailment from MN to PN collapses on its own terms.⁷

Defenders of MN offer several responses. Pennock argues that the distinction between Duhemian and strong MN is less sharp than Plantinga supposes: MN is not merely a convenient heuristic that could in principle be abandoned while leaving science intact; it is constitutive of the practice of science in the same way that the requirement for logical consistency is constitutive of mathematics.⁴ Boudry, Blancke, and Braeckman argue that the question of whether MN is an empirical discovery or a prior methodological commitment is a false dilemma: it can be both, justified by its track record and refined in light of continued success.¹⁸ Martin Mahner contends that Plantinga’s distinction, while philosophically precise, does not generate a practical alternative: no scientist has successfully operationalised a method for detecting supernatural causation that could be submitted to peer review, replicated, or applied consistently.⁹

MN and the god-of-the-gaps problem

The relationship between methodological naturalism and god-of-the-gaps reasoning is direct and important. The god-of-the-gaps pattern — inferring divine action from gaps in scientific knowledge — is exactly what MN is designed to prevent, not because gaps are impossible but because filling them with supernatural agency is scientifically inert. Once a gap is filled with “God did it,” the incentive to seek a naturalistic mechanism disappears, and the problem is dissolved rather than solved.^{5, 10}

Stephen Jay Gould’s non-overlapping magisteria (NOMA) framework, presented in Rocks of Ages (1999), approaches the same territory from a slightly different angle. Gould argued that science and religion occupy distinct magisteria, or domains of teaching authority: science covers the empirical realm of fact and theory, religion covers questions of ultimate meaning and moral value, and the two do not genuinely conflict because they do not address the same questions.¹⁰ NOMA entails MN as a corollary: if religion’s domain does not include empirical claims about the natural world, then invoking divine action to explain natural phenomena is a category error — it crosses a boundary that both science and good theology should respect. Critics of NOMA, including Richard Dawkins, have argued that it concedes too much by suggesting that religion has no empirical content at all; but even critics who reject NOMA as a complete account of the science–religion relationship typically accept its entailment that the scientific investigation of nature should proceed naturalistically.²⁰

The asymmetric track record of gap-filling is the most powerful practical argument for MN and against gap-based reasoning. Every phenomenon once attributed to supernatural causation that has subsequently received a scientific explanation has received a naturalistic one. The motion of planets was attributed to divine sustenance; it was explained by gravitation. The diversity of species was attributed to separate divine creation; it was explained by natural selection. The stability of the genetic code was thought to require a special vital force; it was explained by molecular biology. Disease was attributed to spiritual causes; it was explained by microbiology. In no case has a phenomenon previously explained naturalistically subsequently required the re-introduction of a supernatural explanation. The flow is entirely one-directional.^{8, 13} This pattern does not prove that no phenomenon will ever turn out to require a supernatural explanation, but it provides a strong inductive warrant for betting that current gaps will close on naturalistic terms — and for treating the methodological discipline of seeking natural explanations as the most productive available strategy.

Scope and limits

Defenders of MN are careful to distinguish what it does and does not imply. It does not imply that science has explained everything, or that everything is in principle explicable by current science, or that the questions humans care most about — the meaning of suffering, the basis of moral obligation, the nature of consciousness, the existence of God — are answerable by scientific methods.^{4, 19} MN is a constraint on scientific explanation, not a claim about the total scope of human knowledge. Philosophy, literature, theology, and personal experience address dimensions of human life that fall outside science’s domain, and MN does not deny this.

What MN does deny is that science can adjudicate claims by appeal to supernatural agents. A cosmologist can describe what happened in the first moments after the Big Bang with extraordinary precision; she cannot, as a scientist, say whether God chose those initial conditions. A biologist can describe in molecular detail how natural selection shapes protein structure; he cannot, as a biologist, say whether God guides mutations. This is not a deficiency of science; it is a recognition that the instruments of science are calibrated for the natural world, and deploying them to answer metaphysical questions about the existence or agency of God yields only silence — not a negative answer, but no answer at all.⁴

The demarcation between science and non-science has occupied philosophers of science at least since Karl Popper proposed falsifiability as the criterion. The consensus view in contemporary philosophy of science is that no single criterion is fully adequate, but that a cluster of features — testability, reproducibility, natural causation, peer review, consistency with established knowledge — collectively identifies the scientific enterprise.^{9, 13} MN is one node in this cluster. Removing it while preserving the others is not straightforwardly possible: a science that admits supernatural explanations would face the problem that such explanations compete with natural ones, cannot be tested against them, and yet are capable of “explaining” any observation by varying the attributes and intentions of the posited supernatural agent. The methodological discipline of naturalism is not an optional add-on to science; it is part of what makes scientific explanation explanatory rather than merely verbal.

The distinction between MN and PN also means that the success of science under MN does not itself establish philosophical naturalism. Science working naturalistically is consistent with theism: a theist can say that the natural order through which science operates is the natural order God created and sustains. The practical success of naturalistic science from Bacon through Darwin to the present is evidence for MN as a research strategy; it is not, by itself, evidence that God does not exist.¹⁹ This is a point that both defenders of MN and thoughtful critics like Plantinga agree on, even as they disagree about MN’s philosophical status. Where the debate is genuinely contested is not over the practical utility of MN, which is attested by the entire history of modern science, but over whether adopting it as a methodological commitment carries covert philosophical implications that its defenders are unwilling to acknowledge.