Pascal's Wager

The argument in plain language

Pascal asks you to imagine that the existence of God is a question reason cannot settle. The cosmologist's instruments give no decisive answer; the metaphysician's proofs are all contested. You must, however, place a bet. Living any life at all is already a bet — you either spend it as if God exists or as if He does not. Refusing to choose is itself a choice.

If you bet on God and God exists, says Pascal, you gain "an infinity of an infinitely happy life." If you bet on God and He turns out not to exist, you have given up a few worldly pleasures — a finite cost. Bet against God: if He does not exist you gain a small finite reward (a few extra Sundays in bed); if He does exist you forfeit infinite happiness and may earn infinite damnation. The expected utility of belief swamps disbelief by an infinite margin, however small the probability of God's existence. Therefore — Pascal concludes — the rational agent believes.

The reasoning is striking because it sidesteps every traditional proof of God's existence. Pascal does not argue that God is real. He argues only that you should act as if He is, and that your reasons for doing so are coldly arithmetical.

The decision matrix, made explicit

Strip the argument to a four-cell payoff table — exactly the form a modern decision theorist would draw — and the structure becomes vivid:

Your choice	God exists	God does not exist
Believe	+∞ (eternal bliss)	−f (small finite cost: missed pleasures, observance)
Disbelieve	−∞ (eternal loss / damnation)	+f (small finite gain: free Sundays)

Let p be your subjective probability that God exists. The expected utility of belief is p · ∞ + (1 − p) · (−f) = +∞ for any p > 0. The expected utility of disbelief is p · (−∞) + (1 − p) · f = −∞ for any p > 0. As long as you assign even a vanishingly small non-zero probability to God's existence, belief dominates.

This is the algebraic skeleton modern philosophers focus on. It is also where the trouble starts: any argument that yields the same infinite verdict for almost any input probability has either discovered something profound or quietly broken its premises.

Pascal's Wager vs other arguments for God

	Ontological (Anselm)	Cosmological (Aquinas)	Design (Paley)	Moral (Kant)	Pascal's Wager
Type of argument	A priori	A posteriori	A posteriori	Practical	Pragmatic / prudential
Tries to prove God exists?	Yes	Yes	Yes	Postulates God	No — only that you should believe
Relies on natural facts	No	Causation	Apparent design	Moral law	None
Specifies which God	Maximally great being	Prime mover	Designer	Moral lawgiver	Underspecified (the wager's flaw)
Assumes infinite utilities	No	No	No	No	Yes
Era	1078	1265	1802	1788	c. 1660

Notice how Pascal's argument is the odd one out: every other classical argument tries to show that God actually exists. Pascal alone says the question of existence may be unresolvable, and recommends belief anyway on the grounds that it is the optimal hedge.

Worked example: a sceptic at the salon

Suppose Pascal sits across from a 17th-century sceptic — let us call her Madame de Sablé — who tells him she assigns a probability of just 1 in 1,000 to the Christian God's existence. She thinks the case for God is poor; she finds the resurrection implausible; she has read Montaigne. Pascal does not flinch.

"What do you take eternal happiness to be worth?" he asks. She protests that the question is silly because eternity has no price. "Then we agree," he says. "Let us call that quantity ∞. Now: 0.001 × ∞ is still ∞. The expected reward of believing exceeds the expected reward of disbelief by an infinite margin. Why are you still betting against?"

The sceptic's natural reply — that she cannot just decide to believe — Pascal had already prepared for. His advice is therapeutic, almost behaviourist: kneel, take holy water, attend mass, surround yourself with believers. Habit will do the work that argument cannot. "This will naturally make you believe, and deaden your acuteness."

This is a remarkable line. Pascal is not pretending the act of will produces sincere belief immediately. He is recommending a regimen of practices on the grounds that beliefs follow conduct as smoke follows fire. Whether or not the recipe works, it is a strikingly modern picture of how minds are shaped — closer to William James and modern habit psychology than to the rationalist tradition Pascal himself inherited from Descartes.

The standard objections

The many-gods objection

The most damaging response, due in its modern form to Denis Diderot in the 18th century and sharpened by William James and J. L. Mackie, runs like this. The Christian God is one option among many. Allah, Vishnu, Zeus, the demanding god of some unborn religion, and infinitely many possible deities each promise infinite reward to their devotees and infinite punishment to apostates. Several of them — most strikingly the jealous monotheisms — punish you for worshipping a rival.

If two gods both yield expected utility +∞, expected-utility reasoning cannot choose between them. The wager points equally to every faith and so picks none. Worse: a god who rewards honest atheists and damns those who believe out of self-interest is just as logically possible as Pascal's Christian God, and would invert his recommendation.

You cannot choose to believe

Bernard Williams called this the problem of "doxastic voluntarism." Belief, on his view, is involuntary: I cannot decide to believe that there is a giraffe in my kitchen any more than I can decide that the sky is green. Pascal's habit-based reply may produce outward conformity but, the objection goes, what an infinite God presumably cares about is the genuine state of your soul — and a habituated belief produced for prudential reasons is exactly the kind of insincere worship a discerning deity might condemn.

Infinite utilities break the math

Decision theorists since Alan Hájek have noted that infinite expected utilities create paradoxes. If belief has expected utility ∞, then so does belief plus a coin-flip; so does belief minus stealing a loaf of bread. Mixed strategies that rationally shouldn't tie with sincere devotion end up tied at +∞. Strict expected-utility maximization ceases to discriminate between morally and prudentially distinct actions. The infinity that gives the wager its rhetorical force also dissolves it.

The cost may not be finite

Pascal assumes the worldly cost of belief is small. A modern critic — say, the New Atheists or the secular humanist tradition — points out that a life lived under religious constraint can carry significant costs: hours of observance, lost autonomy, alienation from family, sometimes participation in harmful institutions. If the finite "f" in the matrix is itself large, the argument loses some of its asymmetry — though, granted, any finite number is still dominated by ∞.

Variants and modern descendants

• James' Will to Believe (1896). William James argued that for "live, forced, momentous" choices that reason cannot resolve — including the religious one — we are entitled to let our passional nature decide. James drops the infinite-utility math and rehabilitates the wager as a license, not a proof.
• Pragmatic Wagers in Climate and AI Policy. Modern decision theorists deploy Pascal-shaped arguments in cases of catastrophic risk: the expected loss from runaway climate change or unaligned AGI may justify costly precautions even if their probability is small. Critics call these "Pascal's Mugging," after Eliezer Yudkowsky's name for the way infinite-stakes arguments can hijack rationality.
• Cromwell's Rule. The statistician Dennis Lindley argued that you should never assign probability exactly 0 or 1 to any contingent claim. This makes Pascal-style arguments hard to dismiss by setting p(God) to zero.
• Atheist's Wager. Moritz Schlick and others have flipped Pascal: live virtuously without belief; if God exists and is good, He rewards virtue; if not, you have lived well anyway. This neutralises the asymmetry by denying that belief itself is what God values.
• Quantum / many-worlds variants. Some 21st-century writers have considered analogous wagers across branches of a many-worlds universe, where the "infinite" payoff becomes a sum across futures rather than across an afterlife.

Pascal in his moment

Pascal was, by 1660, the inventor of probability theory (in correspondence with Pierre de Fermat about a gambler's problem in 1654), the builder of the first mechanical calculator, and a contributor to the science of vacuums and hydraulics. He was also a Jansenist — a member of the austere Augustinian movement based at Port-Royal — and convinced that the natural human condition was misery without grace.

The Pensées were notes for a never-finished Apology for the Christian Religion. The fragment we now call the Wager was meant to crack the shell of the worldly Parisian sceptic and make him willing to take the next step into religious practice. Pascal did not think the Wager was faith. He thought it was the staircase up to it. Fragment 277 is the famous reminder that "the heart has its reasons, of which reason knows nothing."

Common confusions

• The Wager is not a proof of God's existence. Pascal was explicit that reason cannot settle the question. He offers a decision rule under uncertainty, not an argument for theism in the traditional sense.
• Pascal did not invent decision theory in order to defend Christianity. The mathematical apparatus was real and original — modern game theorists still cite the Pensées — but it grew out of his work on gambling problems, not his apologetics.
• "Acting as if you believe" is not the same as believing. Pascal's habit-based prescription is a mechanism for cultivating belief over time, not a claim that pretending counts.
• The infinite payoff is doing all the work. If you replace +∞ with any finite number, however large, the wager collapses unless you also know p(God) is high enough. Critics see this load-bearing infinity as the argument's structural weakness.
• The Wager does not specify which God. Pascal himself wrote with the Christian God in view, but the argument as stated points at any infinitely-rewarding deity — which is exactly the many-gods problem.