The setup
In 1935, Einstein, Podolsky, and Rosen (EPR) published an argument that quantum mechanics must be incomplete. They considered pairs of particles prepared in entangled states — states where measurements on one particle predict outcomes of measurements on the other, even if the particles are far apart.
Their reasoning: if measuring particle A on Earth instantaneously tells you something about particle B on Mars, then either:
- The particles had definite properties all along (we just didn't know them — "hidden variables"), or
- Some faster-than-light influence is happening between them.
EPR considered faster-than-light influence absurd (rightly, given relativity). Therefore, they concluded, particles must have had definite properties all along, and quantum mechanics — which doesn't include such properties — is incomplete. There must be deeper "hidden variables" we don't have access to.
This was a coherent and serious philosophical challenge to quantum mechanics. It seemed for decades that EPR's question was unresolvable — a matter of metaphysical preference rather than physics.
Until John Bell.
What Bell proved (1964)
John Bell, working at CERN in 1964, proved a theorem of stunning generality:
Any local hidden-variable theory must satisfy certain mathematical inequalities for correlations between distant measurements. Quantum mechanics predicts these inequalities are violated.
Specifically, Bell considered any theory satisfying two assumptions:
- Realism: physical systems have definite properties even when not being measured.
- Locality: the outcome of a measurement on one system cannot depend on the choice of measurement made on a distant system (specifically: no influence propagates faster than light).
Bell showed that under these assumptions, certain statistical correlations between distant measurements are bounded by specific inequalities (now called Bell inequalities). Quantum mechanics predicts that these bounds are violated by entangled particles.
So Bell turned the EPR question from metaphysics into experimental physics: measure the correlations. If Bell inequalities are satisfied, EPR were right and quantum mechanics is incomplete. If Bell inequalities are violated, EPR were wrong and local realism is false.
A simplified version of Bell's argument
The mathematical details vary, but here's a flavor:
Imagine two particles entangled in a specific way, flying apart to two distant detectors A and B. Each detector can be set to one of several measurement angles (let's say θ_A for detector A and θ_B for detector B). Each detector outputs +1 or -1 for each particle.
The correlation is the average of the products A × B over many trials.
A local hidden-variable theory says that each particle has some hidden property λ that determines its response to any possible measurement angle. The correlation between A and B depends on the joint statistics of λ and the chosen measurement angles.
Bell showed that for any local hidden-variable theory, certain combinations of correlations at different angle pairs must satisfy specific inequalities. The simplest is the CHSH inequality (Clauser, Horne, Shimony, Holt, 1969):
|E(a,b) - E(a,b') + E(a',b) + E(a',b')| ≤ 2
where E(a,b) is the correlation when detector A is set to angle a and detector B to angle b.
Quantum mechanics predicts this combination can reach 2√2 ≈ 2.83 — exceeding the local-realist bound of 2. So if you measure these correlations and get values above 2, you've ruled out local realism.
This is a math fact, not a philosophical claim. Either local realism is true (and the inequality holds), or quantum mechanics is right (and the inequality is violated). Both can't be true.
The experimental tests
Bell's theorem opened a path: do the experiment. Measure the correlations and see.
This took some time, because making and measuring entangled pairs is hard, and closing potential loopholes is harder still. The progression:
1972 (Freedman and Clauser, Berkeley): first Bell test, using polarization-entangled photons from atomic cascades. Confirmed quantum mechanics over local hidden variables, but with significant loopholes.
1982 (Aspect, Grangier, Roger, Orsay): improved experiment with faster switching of detector orientations, closing the locality loophole more decisively. Strong confirmation of quantum mechanics.
1998 (Weihs et al., Innsbruck): photon-pair entanglement over 400 m with random fast switching of detectors. Locality loophole effectively closed.
2015 (Hensen et al., Delft; Shalm et al., NIST; Giustina et al., Vienna): "loophole-free" Bell tests closing multiple loopholes simultaneously. Definitive confirmation: quantum mechanics is right; local realism is wrong.
2018-onwards: "cosmic Bell tests" using astrophysical sources (light from distant stars and quasars) to determine measurement settings — closing the "freedom of choice" loophole as completely as physically possible.
2022 Nobel Prize in Physics: awarded to Alain Aspect, John Clauser, and Anton Zeilinger for their pioneering experimental work on entanglement and Bell tests.
The cumulative experimental evidence is overwhelming. Local hidden-variable theories are ruled out. Nature is not locally realist.
What this means
Bell's theorem and its experimental confirmation tells us something deep about the world: at least one of the following must be false:
- Realism: systems have definite properties prior to measurement.
- Locality: no causal influences propagate faster than light.
- Freedom of choice: experimenters' choices of measurement settings are independent of the systems being measured (no superdeterminism).
- The standard probability calculus: classical probability theory applies to nature.
Most physicists conclude that realism, locality, or both must give way. Different interpretations of quantum mechanics handle this differently:
- Copenhagen-ish: properties don't exist before measurement (realism fails).
- Many-worlds: there's no single outcome to be realist about (realism is reinterpreted).
- Bohmian mechanics: explicit non-locality at the level of the pilot wave (locality fails).
- Objective collapse: similar non-local effects from the collapse process.
- QBism: outcomes are agent-experiences, not objective facts (realism is reinterpreted).
- Superdeterminism: the freedom-of-choice loophole isn't really closed because nothing is free; this is a fringe position because it's hard to take seriously without sacrificing science itself.
The most common informal summary: "spooky action at a distance is real." Einstein meant this pejoratively in his original objection to quantum mechanics; Bell's theorem and experimental confirmation showed that, in a specific carefully-defined sense, Einstein's spooky action was indeed how nature works.
But it's not "spooky" in the sense you might think
A crucial subtlety: Bell's theorem doesn't say you can send signals faster than light using entanglement.
The correlations between entangled particles are real and violate Bell inequalities. But the outcome at each individual detector is random (each side sees ~50/50 + and -). You can only observe the correlation by COMPARING the results from both sides — which requires bringing them together at less than the speed of light.
So Alice can't use her entangled particle to send a message instantly to Bob's entangled particle. The "spooky action" doesn't carry usable information. Special relativity remains intact at the level of signaling.
This is sometimes called "peaceful coexistence" between quantum mechanics and special relativity. There's no causal paradox, no FTL communication, no time-loop trouble. But there IS something genuinely non-local about the correlations themselves — they're stronger than any local realistic theory permits.
This is the conceptual subtlety. The world is "non-local" in the correlations between entangled measurements, but "local" in the sense that no signals propagate faster than light. Both statements are true; they don't contradict each other; but they're hard to wrap intuition around.
Why "freedom of choice" matters
One of the more philosophically interesting loopholes is the "freedom of choice" or "superdeterminism" loophole.
Bell tests require that the experimenters' choices of measurement settings are independent of the particles being measured. If somehow the particles "knew" in advance what measurements would be performed, they could appear to violate Bell inequalities without actually being non-local.
This requires a kind of cosmic conspiracy: the laws of physics must be set up so that the source of entangled particles "knows" what measurements the experimenters will choose, and produces particles with appropriately correlated hidden values. The experimenters' apparent free choices would actually be predetermined to match the particles.
Cosmic Bell tests use light from distant astronomical sources (quasars billions of light-years away) to determine measurement settings. For superdeterminism to explain the results, the original quasars and the particle sources today must have been correlated since the early universe — a more extreme form of cosmic conspiracy.
Most physicists consider superdeterminism unattractive because it essentially abandons the assumption that scientific experiments give meaningful information about nature (if the universe is conspiring against us, every experiment is suspect). But the loophole is technically real and can't be fully closed by experiment.
Useful applications
Bell tests aren't just philosophical — they have practical applications:
Device-independent quantum cryptography: by performing Bell tests on the entangled key-distribution photons, you can verify the quantum-mechanical correlations are intact, ruling out a class of attacks where the equipment might be compromised.
Quantum random number generation: Bell-test violations certify true quantum randomness, useful for generating provably random numbers for cryptography.
Tests of quantum gravity: extending Bell tests to involve gravitational interactions might eventually probe whether gravity is quantum-mechanical.
Network entanglement: Bell tests over networks demonstrate that multi-party quantum correlations work as predicted, an important step toward quantum internet protocols.
A note on what Bell didn't prove
Some common misreadings of Bell's theorem:
"Bell proved quantum mechanics is right." Not quite. Bell proved that local hidden-variable theories can't reproduce quantum predictions. Experiments then confirmed the quantum predictions and refuted local hidden variables. Bell didn't prove quantum mechanics directly; he set up an empirical test.
"Bell proved consciousness creates reality." Bell said nothing of the kind. Bell's theorem is about local realism, not about observers or consciousness.
"Bell proved there are no hidden variables." Bell proved no LOCAL hidden variables. Bohmian mechanics is a non-local hidden-variables theory that's consistent with Bell's theorem.
"Bell proved faster-than-light signaling is possible." No — quantum non-locality cannot be used for signaling. Special relativity at the level of information transfer is preserved.
"Bell solved the measurement problem." No — Bell's theorem constrains hidden-variable theories but doesn't resolve what happens during measurement.
If you'd like a guided 5-minute course on Bell's theorem and entanglement, NerdSip can generate one.
The takeaway
John Bell proved in 1964 that any local hidden-variable theory must obey specific mathematical inequalities for correlations between distant measurements on entangled particles. Quantum mechanics predicts these inequalities are violated. Experiments since 1972 — culminating in loophole-free tests since 2015 and the 2022 Nobel Prize for Aspect, Clauser, and Zeilinger — have confirmed the violations to extraordinary precision. The conclusion: nature is not locally realist. Either properties don't exist before measurement, or non-local correlations (in a specific weak sense that doesn't permit faster-than-light signaling) really happen, or both. This is one of the most profound experimental results in modern physics, transforming a metaphysical debate into a settled empirical fact — even if its full implications remain philosophically debated.