The Wager Against Reality: Bell's Theorem and the End of Local Realism

The Hook: When Metaphysics Moved Into the Lab

There is a rare kind of statement in science that accomplishes something seemingly impossible: it takes a question philosophers have argued over for centuries without result and turns it into an experiment you can set up in a laboratory on a single afternoon and settle for good. Bell's theorem is the most famous of these statements. It transformed the question "Does the world exist independently of whether we observe it?" — a question one might have thought hopelessly vague and forever undecidable — into a plain inequality between measured numbers.

The physicist Henry Stapp once called Bell's theorem "the most profound discovery of science." That may sound grandiose, but the reason behind it is precise. Almost all the great statements of physics tell us how nature behaves. Bell's theorem tells us something rarer and more radical: it tells us what nature cannot be. It rules out an entire class of conceivable worlds — namely, all worlds that are at once "real" and "local," in a sense to be made exact in a moment. And the astonishing part is this: precisely the kind of world Albert Einstein considered the only reasonable one belongs to the excluded class.

On 4 October 2022, Alain Aspect, John Clauser and Anton Zeilinger were awarded the Nobel Prize in Physics — "for experiments with entangled photons, establishing the violation of Bell inequalities and pioneering quantum information science." With this the Nobel committee honored not merely three experimentalists but the conclusion of a nearly sixty-year story: the path from a philosophical quarrel about the nature of reality to an experimentally settled fact — and on to a technology built on precisely that fact.

This article takes you the full distance: from Einstein's objection of 1935, through Bell's ingenious reformulation of 1964, the measurable CHSH inequality, the decisive experiments, the laborious closing of the last "loopholes," to the surprisingly concrete applications in cryptography and IT security. By the end it should be clear why a result that looks like pure metaphysics has today become the foundation of a new security technology.

Part 1: The Dispute – Einstein, EPR, and the Completeness of Quantum Mechanics

Two Assumptions That Seemed Self-Evident

To understand Bell's theorem, one first has to understand exactly what it excludes. The point is the combination of two assumptions that, to Einstein — and honestly, to most people with common sense — seemed non-negotiable.

The first assumption is realism (more precisely: value realism, or "hidden variables"). It holds that physical properties possess a definite value before and independently of being measured. The moon is there even when no one is looking. An electron has a spin, a photon a polarization, and measurement merely reads off this pre-existing value. If quantum mechanics can only give probabilities, then — so the realist argues — only because our description is incomplete. There are "hidden variables" that in truth fix the outcome; we simply do not know them.

The second assumption is locality. It holds that no event can instantaneously influence another across arbitrary distances. Effects need time to propagate, and nothing outruns light. What I set here and now on my measuring device cannot alter, in that same second, the result of an experiment a light-year away. This assumption is the heart of Einstein's special theory of relativity.

The conjunction of both assumptions is called local realism. It is the worldview that things have fixed properties and that action at a distance is ruled out. It is precisely this worldview that Bell's theorem will ultimately bring down.

The EPR Argument of 1935

In 1935 Albert Einstein, Boris Podolsky and Nathan Rosen published a famous paper with the provocative title "Can the Quantum-Mechanical Description of Physical Reality Be Considered Complete?" Their answer was a firm no. The argument — known today as the EPR paradox — is a masterpiece of logic.

Consider two particles that originate from a common source and whose properties are correlated with one another (today we would say: entangled, a term Erwin Schrödinger coined that same year). From conservation laws one knows, for instance, that their spins must be opposite: measure "up" on particle A, and B is guaranteed to be "down." Now carry the two particles far apart. If you now measure A, you instantly know the result for B — without touching B at all.

Einstein, Podolsky and Rosen argued: if I can predict the value of B with certainty without disturbing B, then that value must be an "element of physical reality" — it must objectively have been there already. And since I can freely choose which property of A to measure, and thereby predict different properties of B, all these properties of B must be simultaneously real. But quantum mechanics cannot ascribe them sharp values at once (Heisenberg's uncertainty relation forbids it). Therefore, EPR concluded, quantum mechanics is incomplete. There must be a deeper description — with hidden variables — that fixes everything.

Niels Bohr replied that same year, but his rejoinder was more philosophical than mathematical and by no means convinced everyone. For nearly three decades the debate was regarded as a matter of interpretation, of personal taste, of metaphysics — at any rate as something that could not be settled in the laboratory. You could be an Einsteinian or a Bohrian, but you could not, it seemed, refute either. This is precisely where the breakthrough begins.

Part 2: Bell's Ingenious Idea of 1964

A Northern Irishman Who Wanted Einstein to Be Right

John Stewart Bell, born in Belfast in 1928, was a theoretical physicist who actually worked on particle accelerators at CERN. He pursued the foundational questions of quantum mechanics, as he put it himself, as a "hobby," on the side, but with great seriousness. Remarkably, Bell sympathized with Einstein. He found hidden variables attractive and secretly hoped to help them prevail. That he ended up proving the exact opposite makes his intellectual honesty all the more impressive.

In 1964 Bell published a short paper titled "On the Einstein Podolsky Rosen Paradox" in the obscure, short-lived journal Physics Physique Физика. The paper is only a few pages long and contains hardly more than elementary probability theory. And yet it changed physics. Bell's decisive insight was this: if you take local realism truly seriously — if, that is, there are hidden variables that fix the outcomes locally — then quantitative limits follow for the strength of the correlations one can observe between two distant measurements. And these limits differ measurably from what quantum mechanics predicts.

That is the stroke of genius. EPR had believed that hidden variables and quantum mechanics were empirically indistinguishable — that hidden variables were merely a nicer story behind the same predictions. Bell showed this to be false. Every local-realistic theory — no matter how complicated, no matter which hidden variables it assumes — obeys an inequality. Quantum mechanics violates it. With that the question had become decidable.

Why There Must Be a Limit – a Mental Picture

The deeper reason for Bell's inequality can be understood without formulas. Imagine a source sending one particle each to two distant stations — call the experimenters Alice and Bob. Each station can choose one of several measurements on its particle (say: testing polarization at various angles), and each measurement yields one of two outcomes, which we encode as +1 and −1.

In the local-realistic picture, every particle carries a kind of invisible "cheat sheet" that already specifies, at the moment of creation, how it will answer every possible measurement. The two cheat sheets may be correlated with one another (they do come from the same source), but once the particles are separated, Alice's choice of measurement can no longer alter the cheat sheet of Bob's particle — locality forbids it. Bell then showed, by simply counting through all possible cheat sheets: no matter how the source distributes these sheets, the correlations between Alice's and Bob's results cannot exceed a certain limit. At heart it is a combinatorial argument, akin to noting that in a survey the fractions of people who like A and B, B and C, and A and C cannot be entirely free but must hang together.

Quantum mechanics does not obey this limit, because its correlations do not come from a shared cheat sheet but from entanglement itself — from a state that describes the two particles as an inseparable whole and assigns neither of them fixed values in advance. Anyone wishing to understand the entangled state from the ground up will find the details in the cross-reference Spooky Action at a Distance: Quantum Entanglement from Einstein to the Quantum Internet.

Part 3: The CHSH Inequality – the Measurable Form

From Bell's Original to the Laboratory Version

Bell's original inequality of 1964 assumed perfect anticorrelations — an idealization never exactly attainable in a real laboratory with imperfect detectors. In 1969 four physicists — John Clauser, Michael Horne, Abner Shimony and Richard Holt — generalized Bell's idea into a form robust enough for real experiments. This CHSH inequality (after the initials of its authors) remains to this day the workhorse of all Bell tests.

The construction is elegant. Alice chooses between two measurement settings, call them \(a\) and \(a'\). Bob likewise chooses between two settings, \(b\) and \(b'\). Each measurement yields +1 or −1. For each combination of settings, one measures, over many particle pairs, the correlation value \(E(a,b)\) — essentially the average of the product of the two outcomes. It runs from +1 (the results always agree) through 0 (no relationship) to −1 (the results are always opposite).

From these four correlation values one forms a single figure:

$$S = E(a,b) - E(a,b') + E(a',b) + E(a',b')$$

The entire debate between Einstein and quantum mechanics can now be condensed into the value of this one number \(S\).

Two Numbers That Decide Everything

Here lies the full force of the theorem, in two plain bounds.

First: every local-realistic theory — every world with hidden variables and without action at a distance — must satisfy

$$|S| \leq 2$$

This is the Bell-CHSH bound. It follows inescapably from the mere assumption that the results are fixed in advance and that Alice's choice does not influence Bob's particle. No classical theory, however ingenious, can exceed it.

Second: for suitably chosen measurement angles on a maximally entangled pair, quantum mechanics predicts a value of up to

$$|S| = 2\sqrt{2} \approx 2.828$$

This upper bound is called the Tsirelson bound (after Boris Tsirelson, 1980). It lies noticeably above 2. The region between 2 and 2.828 is, in a sense, no man's land: values there are unreachable for any local-realistic world, yet routine for the quantum world.

The following table summarizes the three relevant regimes:

The crucial point: if an experiment measures a value of \(S\) above 2, then local realism as a whole is refuted. Not one particular theory, but an entire family of conceivable theories falls in a single stroke. That is exactly what the experiments did.

Part 4: The Experiments – From Homebrew Setup to Nobel Prize

Freedman and Clauser, 1972

The first serious Bell test was achieved in 1972 by Stuart Freedman and John Clauser at the University of California, Berkeley. Clauser, then a young and unconventional physicist, had taken up the topic against the explicit advice of established colleagues — foundational questions of quantum mechanics were regarded as a career-damaging pastime. He built a source of calcium atoms that, through an atomic cascade, emitted pairs of polarization-correlated photons, and measured their polarization at various angles.

The result was unambiguous: the data violated Bell's inequality and agreed with quantum mechanics. Einstein's hidden variables took their first empirical blow. But the setup was still crude, and — more importantly — it left loopholes open, which we will turn to shortly.

Aspect, 1982 – the Switching Analyzers

The decisive methodological advance came in 1982 from Alain Aspect and his team in Orsay near Paris. Aspect's setup closed a central loophole, the locality loophole. The objection ran: what if Alice's device somehow — via a slow, unknown signal — "tells" Bob's device which setting was chosen, and the particles then coordinate their answers? As long as the settings are fixed in advance, this can never be entirely ruled out.

Aspect's solution was as simple as it was brilliant: he switched the measurement settings while the photons were already in flight — so fast that between the choice of setting on one side and the measurement on the other, even a light signal would not have had time to transmit the information. This ruled out any possibility that one side could "know" the other's setting. Bell's inequality was violated once again, and more clearly.

Zeilinger and the Maturing of Quantum Information

Anton Zeilinger and his groups in Innsbruck and Vienna refined the technique to perfection from the 1990s onward. They used the by-then-available parametric down-conversion in nonlinear crystals to generate bright, clean sources of entangled photon pairs. Zeilinger closed further loopholes, extended the distances to kilometers, and thereby laid the practical foundations for quantum cryptography and quantum teleportation. In his hands the Bell test gradually became a tool — no longer merely an experiment about the nature of reality, but the starting point of a new engineering discipline.

For these three — Clauser as pioneer, Aspect as the man who attacked the locality loophole, Zeilinger as the one who founded quantum information science — the 2022 Nobel Prize was awarded.

Part 5: The Loopholes and Their Final Closing

Why a Positive Result Was Not Immediately Enough

A skeptic — and in foundational physics skepticism is a virtue — could, even after Aspect's experiments, still object: "You violated the inequality, but perhaps only because your setup left a back door open through which a local-realistic explanation slips back in after all." These back doors are called loopholes. Three are especially important.

The detection loophole: early detectors captured only a small fraction of the photons. If one assumes that the missed photons would have behaved systematically differently from the measured ones, a violation could be faked. One must therefore actually detect a sufficiently high fraction of all pairs (otherwise the "fair-sampling assumption" fails).

The locality loophole: the measurement settings and results on both sides must be space-like separated — far enough apart in space and time that no signal traveling at the speed of light could get from one to the other. Aspect attacked it but could not close it entirely.

The freedom-of-choice loophole: the measurement settings must be genuinely freely and independently chosen. If some common factor in the past had determined both the particles' hidden variables and the seemingly "random" choice of settings, the correlation could have been arranged in advance.

The difficulty was that early experiments could only ever close one loophole at a time. A setup with high detection efficiency left locality open, and vice versa. A critic could always point to whichever hole remained open.

2015: The Year of the Loophole-Free Tests

The breakthrough came in 2015, when three independent groups published, within a few months, experiments that closed the detection and the locality loophole simultaneously.

The first was presented by the group around Ronald Hanson and Bas Hensen at TU Delft. Their trick was entanglement swapping between the electron spins of two nitrogen-vacancy centers (NV centers) in two diamonds located 1.3 kilometers apart on campus. Because the actual measurement was performed on the readily readable electron spins, and only those runs counted in which entanglement had demonstrably succeeded, the setup circumvented the detection loophole. The 1.3 kilometers provided the space-like separation. The result: \(S = 2.42 \pm 0.20\) — a violation that rejected local realism with a p-value of about 0.039.

Almost simultaneously, the group around Marissa Giustina and Anton Zeilinger in Vienna, and the group around Lynden Shalm at NIST in Boulder, reported photon-based experiments with highly efficient superconducting detectors. They achieved a statistical significance far beyond any reasonable threshold of doubt (p-values well below 10⁻⁶). With that, local realism was experimentally finished — no longer merely probably false, but refuted under overwhelming weight of evidence.

2016–2018: The Assault on Freedom of Choice

There remained the most stubborn, almost philosophical loophole: freedom of choice. How does one ensure that the choice of measurement settings is not secretly determined by the same past as the particles? Two experiments attacked it in opposite ways.

The Big Bell Test of 30 November 2016 relied on human free will as a source of randomness. Via a web platform, around a hundred thousand volunteers worldwide — called "Bellsters" — contributed, through a video game, a stream of unpredictable bits; in total 97,347,490 binary decisions, distributed to twelve laboratories on five continents. The idea: if humans choose the settings, then a superdeterministic conspiracy mechanism would also have had to predetermine human behavior.

The cosmic Bell test of 2018 (Rauch, Handsteiner, Zeilinger and colleagues) pushed the source of randomness as far into the past as physically possible: it chose the measurement settings from the color of light from high-redshift quasars, whose photons were emitted billions of years ago. Thus any common factor that had fixed both the quasars and the experiment would have to lie at least 7.8 billion years in the past. The result violated the inequality by 9.3 standard deviations (p ≲ 7.4 × 10⁻²¹) and excluded local-realistic influences from roughly 96 percent of the space-time volume in the experiment's past light cone — effectively from shortly after the Big Bang to today.

The following table gathers the loopholes and how they fell:

Part 6: What Remains? The Escape Routes and Their Price

Something Expensive Must Be Given Up

Bell's theorem is not a proof that "quantum mechanics is strange." It is stricter: it forces us to give up at least one of several deeply rooted assumptions. Which one to sacrifice remains, to this day, the subject of serious debate — yet none of the options is cheap.

One can give up locality. Then there are genuine, instantaneous connections between distant events. This is exactly the path taken by Bohmian mechanics, an explicitly nonlocal hidden-variable theory that reproduces all the predictions of quantum mechanics. The price is action at a distance, which is hard to reconcile with the spirit of relativity.

One can give up realism — the idea that measurement results merely read off pre-existing values. In the standard Copenhagen reading, a particle simply has no definite value before measurement; the measurement produces the result. This is the position of most practicing physicists.

One can give up freedom of choice and accept superdeterminism — the notion that everything, including the choice of measurement settings, has been fixed since the Big Bang and correlated with the particles. This option formally saves local realism, but at the price of undermining the foundation of all experimental science: if nature predetermines our measurements "conveniently," no observation is trustworthy anymore. I am of the opinion that superdeterminism, while not logically refutable, is scientifically barren, because it undercuts the very idea of independent evidence.

Finally, there are interpretations such as the many-worlds interpretation, in which every measurement realizes all outcomes at once and the notion of a single definite result is abandoned. This too carries a high metaphysical price. Anyone wishing to think further about whether a machine or a formalism "really" describes the world will find related reflections in the cross-reference The Chinese Room: Searle and the Question of Whether Machines Can Understand and, in the context of the classical limit, in Why the World Turns Classical: Decoherence, Einselection, and Quantum Darwinism.

The Crucial Misunderstanding: No Faster-Than-Light Communication

A widespread error holds that violating Bell's inequality permits communication faster than light. This is false, and the reason is important. True, the results at the two stations are more strongly correlated than classically possible — yet taken on its own, each side sees only a random sequence of +1 and −1. Only when Alice and Bob compare their records afterward (and this comparison requires a classical channel, hence the speed of light) does the spooky correlation come to light. This property is called the no-signaling principle: quantum mechanics is exactly nonlocal enough to violate local realism, but not so nonlocal as to break relativity. This fine balance is one of the subtlest and most beautiful features of nature.

Part 7: Why This Matters in Practice – From Metaphysics to Security

Up to here one might take Bell's theorem for a magnificent but inconsequential monument. The opposite is the case. Precisely because the Bell violation is so experimentally robust, it has become the foundation of a new security technology.

The key term is device-independent cryptography. In classical quantum key distribution (such as the BB84 protocol), one must trust one's own devices — that the photon detector really does what the manufacturer promises, and has not, say, been tampered with by an attacker. Bell's theorem allows this trust to be replaced by physics. If two parties demonstrate a genuine Bell violation in their data, then the mathematics guarantees that their correlations cannot have been fixed in advance or eavesdropped by a third party — regardless of how the devices are built inside. The Bell violation becomes the touchstone for genuine, non-predetermined randomness.

Two concrete applications follow. First, certified random number generation: one can generate random numbers whose unpredictability is vouched for by a law of nature rather than by trust in a chip — invaluable for cryptography, lotteries, or simulations. Second, device-independent key distribution, whose security holds even if the hardware comes from an untrusted source. For anyone concerned with IT security, this is a remarkable thought: security no longer rests on assumptions about the manufacturer, but on the experimentally guaranteed impossibility of local-realistic explanations.

With this, a circle closes back to the broader upheaval that quantum technology means for cryptography. The same entanglement that refutes local realism powers the quantum computers that may one day break today's encryption — the reason post-quantum cryptography is already being prepared today, as described in the cross-reference Harvest Now, Decrypt Later: Post-Quantum Cryptography and the Race Against the Quantum Computer. And the same fragile entanglement that makes Bell tests so hard is the raw material that quantum error correction must laboriously protect so that usable quantum computers can arise at all — see Order from Noise: Quantum Error Correction and the Road to a Fault-Tolerant Quantum Computer.

The Central Takeaway

The central lesson of Bell's theorem reaches far beyond physics, and in a single sentence it is this: a question that looks like pure metaphysics can become a measurable quantity through the right reformulation. Einstein and Bohr argued for thirty years without any experiment seeming possible. Bell showed that the apparently undecidable question had a testable consequence — a number that lies either below or above 2.

The practical prompt that follows is transferable: the next time you face a dispute that seems to be "merely a question of interpretation" or "pure philosophy" — whether in the architecture of a system, in a security assumption, or in a design decision — do not immediately ask "Who is right?" but rather "What different, measurable consequence would the competing positions have?" Often a supposedly metaphysical quarrel turns out to be empirically decidable, once you phrase it sharply enough. And where no such distinguishing consequence can be found at all, that too is a valuable insight: then it may really be only a matter of taste, and you can save the energy. Bell's genius did not lie in guessing the right answer — he was, after all, hoping for the wrong one — but in posing the question so that nature could answer it.

Reflection Question

Bell hoped to prove Einstein right and proved the opposite — precisely because he honestly submitted the question to experimental decision rather than interpreting it in his own favor. Where in your own work do you cling to an assumption you consider "obviously true" without ever having formulated a test that could refute it — and would you be willing to let it go if the result came out above the line?

Cross-References in the Vault

• Spooky Action at a Distance: Quantum Entanglement from Einstein to the Quantum Internet – the phenomenon of entanglement on which every Bell test rests.
• Why the World Turns Classical: Decoherence, Einselection, and Quantum Darwinism – why the spooky correlations vanish in everyday life and the world appears classical.
• Harvest Now, Decrypt Later: Post-Quantum Cryptography and the Race Against the Quantum Computer – the flip side of quantum technology for today's encryption.
• Order from Noise: Quantum Error Correction and the Road to a Fault-Tolerant Quantum Computer – how the fragile entanglement is technically tamed.
• The Chinese Room: Searle and the Question of Whether Machines Can Understand – the related question of what "really" goes on behind a formalism.

Sources

• The Nobel Prize in Physics 2022 – Press Release, NobelPrize.org: https://www.nobelprize.org/prizes/physics/2022/press-release/
• The Nobel Prize in Physics 2022 – Popular Science Background, NobelPrize.org: https://www.nobelprize.org/prizes/physics/2022/popular-information/
• Bell's Theorem, Stanford Encyclopedia of Philosophy: https://plato.stanford.edu/entries/bell-theorem/
• Loopholes in Bell test experiments, Wikipedia: https://en.wikipedia.org/wiki/Loopholes_in_Bell_test_experiments
• Hensen et al., "Loophole-free Bell test using electron spins in diamond" (Scientific Reports 2016, follow-up analysis to the 2015 Nature experiment): https://www.nature.com/articles/srep30289
• Rauch et al., "Cosmic Bell Test Using Random Measurement Settings from High-Redshift Quasars", Phys. Rev. Lett. 121, 080403 (2018): https://link.aps.org/doi/10.1103/PhysRevLett.121.080403

← All articles