A necessitarian reply to Joe Schmid’s cumulative case for theism
Joe Schmid’s seven-hour cumulative case for theism is perhaps the best popular presentation of the recent philosophical literature on the subject. It is not a pile of gaps arguments. It has a clear Bayesian spine: theism has a respectable prior because it is comparatively compact, uniform, non-arbitrary and coherent; it predicts a wide range of otherwise surprising facts better than naturalism does; and those likelihood advantages accumulate until theism becomes probable.
Nearly all of it can be granted. The conclusion still does not follow. The relevant alternative is not brute contingent naturalism, which faces serious explanatory problems, nor any of the four alternatives Schmid considers. It is necessitarianism: the view that reality never had a range of metaphysical options from which one was selected. Ultimate reality exists necessarily, its intrinsic structure is necessary, and everything downstream follows necessarily from that structure. What appears to be a vast field of unrealized alternatives is instead conceptual, epistemic or counterfactual structure rather than a collection of genuine metaphysical rivals that lost a competition.
This is not an argument that theism is false. Much of the pressure falls on a popular form: a necessary God contingently selects one creation from a range of genuinely possible alternatives. Necessitarian theism remains available: if God necessarily produces the actual order, there is no contingent selection event, and much of Schmid’s explanatory apparatus survives intact. The narrower claim is that his cumulative case does not yet distinguish contingent-creation theism from an important family of necessitarian competitors, and that several of his largest claimed likelihood advantages depend on modal spaces and probability measures that have not been adequately justified.
This matters. Schmid’s case begins by placing enormous pressure on unexplained contingency. The question is whether that pressure is followed all the way to the top.
What can be granted
Cumulative Bayesian reasoning is the right framework. Evidence need not entail a hypothesis to favor it, and a hypothesis can gain substantial support from a pattern of individually modest facts. Priors matter as well, and a theory that merely fits the observations can still be inferior to a simpler, less arbitrary rival. Schmid’s example of a law saying that nature remains uniform until 2050 and then changes makes the point cleanly: empirical fit is not everything.
The world is also striking in many of the ways he emphasizes. Lawfulness calls for attention. Nomological harmony is puzzling. Fine-tuning deserves explanation rather than dismissal. Psychophysical harmony may be the most interesting item in the case. Natural selection explains why certain physical behaviors are adaptive, but on some views it does not explain why phenomenal qualities, reasons, behavior and physical states line up as felicitously as they do. Under epiphenomenalism, selection cannot discriminate among phenomenal mappings that leave behavior unchanged. Under interactionist models, mappings might be altered together with their causal roles while preserving behavioral success. And even if psychophysical identities are necessary a posteriori, alternatives remain epistemically live before the identities are discovered. Anyone who thinks evolution dissolves that problem has not understood it.
Unification is an epistemic virtue too. A hypothesis capable of explaining several apparently disparate facts through one underlying principle has an advantage over a collection of unrelated local explanations.
Those concessions sharpen the central question. What kind of ultimate hypothesis provides the best unification?
First: the ascent and the terminus do not use the same principle
Schmid’s contingency argument begins with a contrast. Five mysterious shiny spheres in a forest call for explanation. If each was produced by an earlier sphere, the problem has not disappeared. Even an infinite regress leaves a question standing: why this series rather than a different series, shiny cubes, or nothing at all.
Notice what supplies the force. The puzzle is not that the spheres are unexplained outright, since by hypothesis each is explained by its predecessor. The puzzle is that an alternative was available and did not obtain. The general principle he draws from the case has the same shape: both the existence and the non-existence of a contingent thing are genuine possibilities, only one obtains, and that asymmetry demands an account. The demand is contrastive.
Suppose the reasoning succeeds and contingent reality ultimately depends on something necessary. Call the necessary foundation G and the actual world W. The original question returns:
Why W rather than W?*
One possibility is that the complete necessary reality, including all of its relevant reasons and features, fully determines W. But if the complete necessary basis is sufficient for W, the pressure runs toward W’s necessity. That is the familiar modal-collapse route, and it is the necessitarian result.
A second possibility is that some contingent feature of the divine reality explains the difference: a contingent intention, decree, volition or act. Contingency has then entered the explanatory terminus itself, and the question becomes why that divine state rather than another.
A third possibility is that God’s necessary nature and reasons favor W without determining it. God has adequate reasons for W, perhaps better reasons than for many alternatives, while retaining the power to select W*. That may explain why W was worth choosing and why choosing it was rational. It may even make the selection of W likelier than the selection of many competitors. Unless the reasons settle the outcome, a residue remains.
The fourth possibility is that the choice itself settles it. Schmid says this when he answers the objection that a necessary being cannot explain contingent things without necessitating them: the explanatory link may be indeterministic, with the necessary being freely choosing among outcomes each of which has adequate but non-coercive reasons in its favor.
The reply available to him at this point is that explanation need not be contrastive. The fact that W obtains is explained by God’s reasons together with God’s choice. The contrast between W and W* is not explained: on the version of the principle of sufficient reason he defends, it never needed to be. Nothing is relocated and nothing is left brute.
That is not the theory that powered the ascent. If non-contrastive explanation is good enough at the terminus, it was good enough at the bottom, and the shiny spheres never generated a problem in the first place. Each sphere’s existence was explained by its predecessor. What went unexplained was only the contrast, which is precisely the residue now declared acceptable.
The theist owes an account of the asymmetry. Non-contrastive explanation turns out to be acceptable at exactly one location in the architecture, and that location is where the necessitarian alternative would otherwise win. The license is not a general feature of explanation. It is issued once, by hand, at the summit. The theist rides the PSR as far as the summit and steps out. The principle is indispensable while carrying the argument toward a necessary being and dispensable on arrival.
The necessitarian does not need to prove that brute contingency is impossible. It is enough to observe that contingency has to live somewhere, and that wherever it lives, the explanatory demand that put it there must be switched off for it to survive. (The stronger claim is on the table. Karofsky argues that contingency is an uncontainable contagion: if anything is contingent then everything is, and since not everything is, nothing is. Della Rocca argues that a principle of sufficient reason applied without restriction drives to necessitarianism directly.)
For theists who also hold divine simplicity the pressure sharpens further. Schmid has argued elsewhere that classical theism escapes action-based modal collapse only if the link between God and creation is indeterministic, and has pressed the consequences of that indeterminism for divine intention and providential control. That argument does not bind every theist, since nothing in the cumulative case requires divine simplicity. But for those who hold it, the question of where the difference between producing W and producing W* resides has no comfortable answer.
A theist may simply accept that ultimate free selection is where explanation stops. The debate with necessitarianism then concerns the legitimacy and the cost of that stopping point.
Second: Schmid’s dilemma against necessitarianism is incomplete
Schmid has a sharp reply to the obvious necessitarian response to fine-tuning. Suppose the constants fall within a tiny life-permitting range and the necessitarian says that whatever values the constants possess, they possess necessarily. His answer runs as follows.
The generic hypothesis, that whatever values the constants have they have necessarily, does not predict that the constants will be life-permitting. At the level of epistemic uncertainty it is compatible with any values whatever, so it confers no likelihood advantage. The necessitarian might instead propose that these exact life-permitting values are necessary. That hypothesis entails the evidence, but apparently only because it has been tailored to it. If innumerable alternative values were equally conceivable candidates for being the necessary ones, the specific claim deserves a correspondingly tiny prior.
The first horn is correct. Bare necessity predicts nothing. “It had to be this way” is not an explanation merely because the word has been added.
The second horn does not exhaust the options. A theory can have highly specific consequences without itself being highly specific. What matters for a prior is not how much information appears in the output but how complex the principle generating it is. A short mathematical rule can generate extraordinarily rich structure: the Mandelbrot set is not explanatorily suspicious because its consequences display enormous detail. More generally, deterministic theories entail histories of extraordinary specificity while remaining comparatively compact at the level of fundamental law.
The serious necessitarian proposal cannot be that these constants are necessary because they’re the ones we found. That is ad hoc stipulation and deserves little prior weight. The serious proposal must be that there is an independently motivated, comparatively compact necessary structure from which these values follow. That is a genuine burden and at present it is largely unmet. No developed necessitarian theory has derived the physical and psychophysical structure Schmid emphasizes from a compact independently supported principle.
But it is the right burden, and the corresponding burden on theism should be described at the same level. Bare perfect-being theism does not predict our world in detail either. That a perfect God exists does not by itself tell us to expect these laws, embodied rather than disembodied minds, an evolutionary history, this psychophysical mapping, this distribution of pleasure and pain, this much beauty, or this much hiddenness. Schmid’s likelihood judgments depend throughout on auxiliary assumptions about what a perfect being values, how those values enter creative reasons, which goods receive priority, how trade-offs are handled and how reasons translate into action.
The fair comparison is therefore not between theism and the claim that these constants had to be these constants. It is between families of supplemented theories. A compact necessitarian generator deserves comparison with a compact divine decision theory. Schmid has shown why ad hoc necessitarianism is cheap. He has not shown that every structured necessitarian alternative must be ad hoc.
Third: the prior comparison remains open
Schmid argues that theism receives a respectable prior partly because maximal attributes avoid arbitrary cutoffs. A being powerful to exactly some peculiar finite degree looks less natural than one whose power is unlimited, and similar considerations apply to knowledge and perhaps goodness. One maximally perfect foundation may therefore be simpler than a cosmic committee or a deity with a peculiar collection of quantitative limits.
This does not establish that theism has a higher prior than a comparably compact impersonal necessary structure. The theistic hypothesis still places substantial ideology at the foundation of reality: mentality, intentionality, knowledge, agency, value-responsiveness, perfect goodness and purposive causal power. Perhaps those notions ultimately reduce to one simple perfection principle. That is a substantive metaphysical thesis, not a consequence of calling the resulting being unlimited.
An impersonal necessitarian hypothesis may aspire to comparable compression, and the tradition has candidates, though none is presently in a state to collect. Parmenides and Della Rocca achieve extreme compression by denying the plurality that needs explaining, which forfeits the game against Schmid rather than winning it. Spinoza states the right kind of thesis (infinitely many modes following from the necessity of a substance whose essence includes neither intellect nor will) without ever deriving the physics. Plotinus offers the best shape, an impersonal source producing by necessity with multiplicity downstream. Karofsky argues for the modal thesis and not the compression. The most promising modern resource is ontic structural realism, which makes laws internal to structure rather than imposed on a passive mosaic, and which is not itself a modal thesis at all. If relations are prior to relata there are not two independently specifiable things, laws and contents, that had to be luckily matched, and the rulebook-and-pieces framing of nomological harmony is misdescribed from the start.
Neither side gets simplicity merely by describing itself briefly in English. The relevant issue is theoretical structure. A short phrase can conceal enormous metaphysical machinery, and a mathematically rich theory can rest on a remarkably compact basis. So even before likelihoods are considered, Schmid has shown at most that theism is not prior-prohibitive. The comparison with serious necessitarian rivals has not been made.
Fourth: where do the probability measures come from?
Many of Schmid’s evidential arguments share a recognizable form. There appears to be a vast range of alternatives. Only a small subset is fortunate. Reality occupies that subset. The structure appears in his discussions of constant values, law structures, initial conditions, psychophysical mappings, skeptical scenarios and evolutionary routes to cognition.
The difficulty is that a space of conceivable descriptions is not yet a probability distribution. Conceivability gives candidate descriptions; it does not tell us how much prior weight to assign each one. Where the space is continuous, the observation that the life-permitting interval is tiny is not probabilistically meaningful until a measure has been specified, and changing the measure or the parameterization can change the fraction dramatically.
Schmid knows the crude principle of indifference is problematic, and concludes with Brian Cutter that something in the vicinity of it must be right. The question is whether enough of a principled replacement has been supplied to sustain the very large likelihood ratios sometimes claimed.
A useful test comes from one of his own replies. Against the thought that divine creation becomes improbable because there are vastly more possible micro-arrangements than familiar valuable macroscopic outcomes, he argues that God need not choose uniformly from individual microstates. A divine chooser may attend first to broad kinds: embodied conscious agents, rational life, valuable relationships, beautiful worlds. Raw counting over micro-configurations need not determine the relevant distribution.
The lesson applies symmetrically. Raw counting over conceivable law descriptions, parameter values or psychophysical mappings does not automatically determine the relevant distribution there either.
Schmid has a memorable parity case ready for objections of this family. Imagine the initial conditions of the universe spelling out MADE BY GOD WITH LOVE in every human language. Surely that would be powerful evidence for design. If skepticism about fine-tuning probabilities also eliminates the evidential force of that message, something has gone wrong.
The parity is not exact. The message case still involves probability and is not measure-free. But its low probability under non-intentional generative models is robust across a very broad class of reasonable models. Its structure is highly compressible, semantically specific and closely matched to an independently recognizable communicative intention. One need not impose a single exquisitely chosen measure before the pattern begins to look design-like. The case of a continuous physical parameter is different. Whether a narrow numerical interval is extraordinarily improbable depends much more directly on the prior density assigned across the parameter space, and narrow in one parameterization can become unimpressive in another. A probability argument therefore requires an independently defensible measure, not merely the observation that many numerical values can be written down.
Notice that this is the same distinction that undermined the second horn of the dilemma above: compactness of the generating principle against richness of the output. One technical point is doing both jobs.
None of this makes fine-tuning worthless. It lowers confidence in enormous numerical Bayes factors when the relevant probability space remains poorly understood. The same caution applies to psychophysical harmony. Suppose innumerable alternative mappings between physical and phenomenal states are conceptually describable. It does not follow that they deserve anything like equal epistemic weight. A likelihood argument can certainly run over epistemic possibilities even if only one mapping is metaphysically possible, but it needs a principled distribution over them, and that is the missing step.
Fifth: unification reduces residual surprise
Schmid correctly emphasizes probabilistic dependence on the negative side of the ledger. The problem of evil, divine hiddenness and religious diversity should not automatically be treated as three independent blows. Once horrendous evil has been learned, much of the probability remaining on theism is already concentrated on versions according to which God permits substantial suffering or distance for some reason, so later instances of evil or hiddenness add less negative evidence than they would have in isolation.
That is exactly how sequential conditionalization should work. The same discipline must be applied to the positive evidence, and he never applies it. Suppose we have already learned that reality contains stable laws, highly structured physics, embodied conscious organisms, adaptive behavior and rich social cognition. How much additional surprise remains on discovering that such creatures track behaviorally relevant features of their environment reasonably well, discover regularities in it, distinguish beneficial from harmful states, possess practical forms of reasoning, respond aesthetically to pattern and novelty and symmetry, and avoid many forms of radical skepticism?
These facts need not become evidentially worthless. The point is that their residual likelihood contribution may be far smaller than their isolated presentation suggests. That creates a structural tension in the cumulative rhetoric. Theism’s attraction lies partly in its ability to unify many good-making facts under one broad purposive pattern. But as more of those facts become expected manifestations of one already-learned pattern, the confirmation delivered by each later datum falls.
Unification does not require probabilistic independence, and cumulative evidence does not stop accumulating merely because it is correlated. The narrower point is that a common explanatory source substantially reduces the residual surprise of later evidence.
Some of the evidence survives this better than the rest. Psychophysical harmony may remain a distinct datum if the physical facts leave the phenomenal mapping substantially open, and Schmid’s own conceivability apparatus is designed to establish that. Fine-tuning may remain distinct from the mere existence of laws. But discoverability, non-skepticism, beauty, aesthetic appreciation, practical rationality and several subdivisions of psychophysical harmony look much more dependent once the earlier evidence has been conditioned on. A cumulative case containing twelve labels need not contain twelve large likelihood boosts.
Beauty illustrates the point. A world rich enough to support complex embodied intelligence was already likely to contain symmetry, pattern, contrast, novelty and perceptually salient structure. Creatures adapted to such a world were already likely to possess evaluative responses to their surroundings. Beauty can still favor theism. It should not be treated as a fresh windfall nearly independent of the physical, biological and cognitive facts already in the background. The same symmetry should govern ugliness. If the waste, predation, disease and suffering found in nature are discounted because they belong to a larger integrated order, positive aesthetic data embedded in that same order cannot receive a completely different treatment.
Sixth: religious experience and explanatory defeat
It is tempting for a non-theist to concede this one, and the concession is usually overdrawn. It rests on confusing a thesis about modality with a gap in naturalistic psychology. Necessitarianism says nothing about why human beings have the experiences they have. It inherits whatever psychological, social and cultural account of religious cognition is correct, and there is a great deal of that account already on the table.
Guthrie on anthropomorphism, Barrett on hyperactive agency detection, Boyer and Atran on minimally counterintuitive concepts as byproducts of ordinary cognition, Sosis on costly signaling, Norenzayan on big gods and scaled cooperation, Wilson’s group-selectionist treatment in Darwin’s Cathedral, Whitehouse on modes of religiosity. These accounts disagree sharply with one another, and the adaptation-versus-byproduct dispute is unresolved. The disagreement is about which naturalistic explanation is right, not whether one is on offer. Schmid, describing his atheist years, characterizes religions as sociocultural byproducts thrown up for evolutionary reasons. He never returns to the thought when he makes the argument.
The internal point is the more serious one. Schmid does consider neurological explanation, and answers that every experience has a neural correlate, so pointing to one cannot show an experience delusive. That disposes of crude debunking and misses the cognitive science. The claim there is not that religious experience has a neural substrate. It is that the mechanism producing the experience runs the same way whether or not anything answers to it. Agency detection fires on rustling grass and on the absence of God alike.
Following Korman and Locke, Schmid holds that a belief is defeated when what it rests on is not explained by the facts it purports to be about (which is how he argues that naturalism sinks moral knowledge, since moral facts move no atoms). Agency detection is not explained by facts about God. Applied symmetrically, the principle threatens the evidential force of any class of religious experience produced by such a mechanism.
The theist’s remaining reply is Swinburne’s: natural processes that produce an experience of God are compatible with God having established them. That is stronger than Schmid’s appeal to correlates, because it puts God back into the explanation of the mechanism rather than merely noting that mechanisms exist. But it is structurally a third-factor response, and Schmid rejects third-factor responses in the moral case. His objection to Enoch is that the missing explanatory connection defeats on its own, whatever the reliability turns out to be. If that holds there, it holds here.
The conclusion should be conditional rather than sweeping. Cognitive science has not shown that all religious experience reduces to agency detection, nor that intense mystical and unitive experiences share a single naturalistic explanation, and allegedly corroborated cases raise further questions, though those are exactly where selective reporting does the most work. But wherever an experience class can be explained by mechanisms insensitive to the divine facts it purports to disclose, Schmid’s own epistemic principle threatens its evidential weight.
One caution for the necessitarian. Explanatory defeat is a loaded weapon. An evolutionary account of our modal and mathematical intuitions defeats the necessitarian’s own metaphysics by the same principle, absent some constitutive story about a priori access. That story is more available to someone who holds that rational structure is necessary and that minds are aspects of it than to anyone trying to connect agency detection to God. But it is owed, and it is not written.
What necessitarianism does not get for free
A serious necessitarian response should resist the temptation to turn every explanatory problem into a victory by prefixing “necessarily.”
Bare necessitarianism predicts nothing interesting. The position becomes explanatorily significant only if an independently motivated structure can be given from which important features of reality follow.
Metaphysical necessity does not erase epistemic uncertainty either. If a psychophysical identity is necessary, a thinker who does not yet know the identity can still assign credence across epistemically live alternatives, and Bayesian arguments can operate over them. The dispute concerns the legitimacy of the probability distribution, not the possibility of rational uncertainty.
Nor does necessitarianism automatically explain consciousness, rationality, morality or beauty. A complete theory must show how those belong to or arise from the necessary structure.
The modal burden is substantial. Necessitarianism denies that any complete alternative totality is metaphysically possible, which quantifies over every totality there might have been, and it cannot be established merely by observing that contingentists have explanatory problems. A necessitarian owes an account of counterfactual discourse, modal reasoning, apparent alternatives, and the distinction between metaphysical and merely epistemic possibility.
Contingentism carries one kind of burden: somewhere in the explanatory architecture, genuine modal openness must be housed, explained, probabilistically structured, or owned as primitive. Necessitarianism carries another: it must defend the empty metaphysical foil class and provide a theory of our apparently indispensable modal practices. Neither side is entitled to treat its own burden as invisible.
What remains of the cumulative case
Schmid’s case does not disappear under these qualifications. It changes shape.
Some evidence plausibly favors theism over many forms of naturalism. The purposive appearance of fine-tuning may do so, and so may some forms of psychophysical harmony.
Other evidence favors something broader: a necessary, unified, non-arbitrary, intelligible structure of reality. The existence of a necessary foundation, the absence of arbitrary ultimate cutoffs, lawfulness, global order and mathematical intelligibility all fit that description, and none of them obviously discriminates between a personal and an impersonal necessary foundation.
Some of the positive evidence is overcounted if presented as a long succession of large independent-looking confirmations. Discoverability, non-skepticism, aesthetic appreciation, beauty, practical rationality and several subdivisions of psychophysical harmony retain some force while contributing considerably less once earlier evidence is in the background.
Religious experience is weakened not by my standards but by Schmid’s own, since the explanatory defeat principle he needs against naturalistic moral epistemology cuts against exactly those classes of religious experience that cognitive science can account for.
The remaining pressure runs specifically against the hypothesis of a necessary being who contingently selects. The explanatory demand motivating the cosmological ascent does not stop mattering once God is reached. The closer the ultimate source comes to invariance, simplicity, necessity and rational completeness, the more pressing the question becomes of where the difference between creating this world and creating another actually resides. If the complete difference is grounded in the necessary source, necessity threatens to transmit. If it resides in a contingent divine state, contingency has entered the terminus. If necessary reasons leave several outcomes open and an irreducibly free choice settles the matter, the theory has located primitive contingent selection at the summit, and has stopped demanding of its own foundation what it demanded of the cosmos. That may be the right place to stop. It is still a stop, and it is made at exactly the point where necessitarianism would otherwise win.
Schmid has made a powerful case that brute, indifferent, contingent naturalism is not the only serious explanatory option. What he has not established is that the alternative must be a chooser. The contingency argument drives explanation toward a necessary terminus. The simplicity argument makes that terminus more invariant the better it works. The Bayesian case then asks us to picture a vast modal field and infer a selector from the fortunate region actuality occupies.
He asks why reality landed so often on the fortunate side of modal space. The prior question is whether reality ever had another side to land on.
References and further reading
Necessitarianism, explanation, and theism
Della Rocca, Michael. “PSR.” Philosopher’s Imprint 10, no. 7 (2010).
Della Rocca, Michael. The Parmenidean Ascent. Oxford University Press, 2020.
Karofsky, Amy. A Case for Necessitarianism. Routledge, 2021.
Schmid, Joseph C. “The BEST EVER Cumulative Case for God.” Majesty of Reason, 2026. Video.
Schmid, Joseph C. “The Fruitful Death of Modal Collapse Arguments.” International Journal for Philosophy of Religion 91 (2022): 3–22.
Schmid, Joseph C. “From Modal Collapse to Providential Collapse.” Philosophia 50, no. 3 (2022): 1413–1435.
Cutter, Brian, and Dustin Crummett. “Psychophysical Harmony: A New Argument for Theism.” In Oxford Studies in Philosophy of Religion, vol. 11, 33–71. Oxford University Press, 2025.
Religious cognition and epistemic defeat
Atran, Scott. In Gods We Trust: The Evolutionary Landscape of Religion. Oxford University Press, 2002.
Barrett, Justin L. Why Would Anyone Believe in God? AltaMira Press, 2004.
Boyer, Pascal. Religion Explained: The Evolutionary Origins of Religious Thought. Basic Books, 2001.
Enoch, David. Taking Morality Seriously: A Defense of Robust Realism. Oxford University Press, 2011.
Guthrie, Stewart Elliott. Faces in the Clouds: A New Theory of Religion. Oxford University Press, 1993.
Korman, Daniel Z., and Dustin Locke. “An Explanationist Account of Genealogical Defeat.” Philosophy and Phenomenological Research 106, no. 1 (2023): 176–195.
Norenzayan, Ara. Big Gods: How Religion Transformed Cooperation and Conflict. Princeton University Press, 2013.
Sosis, Richard. “Why Aren’t We All Hutterites? Costly Signaling Theory and Religious Behavior.” Human Nature 14, no. 2 (2003): 91–127.
Swinburne, Richard. The Existence of God. 2nd ed. Oxford University Press, 2004. See especially chap. 13, “The Argument from Religious Experience.”
Wade, Nicholas. The Faith Instinct: How Religion Evolved and Why It Endures. Penguin Press, 2009.
Whitehouse, Harvey. Modes of Religiosity: A Cognitive Theory of Religious Transmission. AltaMira Press, 2004.
Wilson, David Sloan. Darwin’s Cathedral: Evolution, Religion, and the Nature of Society. University of Chicago Press, 2002.
Petronius Jablonski 