The EPR Paradox and Bell's Theorem

A 1935 thought experiment Einstein meant to embarrass quantum mechanics ended up, three decades later, ruling out an entire way the universe could have worked.

Einstein never accepted quantum mechanics. He could not bring himself to believe that the world was, at its deepest level, random; that the act of measuring a particle was what gave it a definite property; that two particles, once entangled, could remain mysteriously coordinated across vast distances. In 1935, with two younger collaborators — Boris Podolsky and Nathan Rosen — he laid out a thought experiment that he believed would finally embarrass the theory. The paper's title was a question: Can Quantum-Mechanical Description of Physical Reality Be Considered Complete? Their answer was no.

What Einstein didn't anticipate was that, three decades later, a quiet Northern Irish physicist named John Bell would turn the EPR argument inside out — and that experiments would then, painstakingly, rule on which side of the disagreement was right. The verdict was not the one Einstein wanted.

The 1935 challenge

EPR begins with a simple demand: any reasonable physical theory should be both real and local.

Realism: the properties of a particle exist before you measure them. The moon is there when you aren't looking. An electron has a definite spin even when no instrument is checking.

Locality: nothing can influence anything else faster than light. Special relativity is built on this. Push a system here, and any effect at a distance has to wait for a signal to arrive.

Quantum mechanics, in the Copenhagen reading Einstein hated, seemed to violate both. A particle had no definite spin until measured; the measurement itself produced the value. And worse, if you had two entangled particles — call them A and B — separated by miles, measuring A's spin would immediately determine what you'd find for B's. As if the measurement on one side reached out and arranged the other.

Einstein called this “spooky action at a distance” and refused to take it seriously. He, Podolsky, and Rosen proposed the obvious alternative. Each particle, they said, must carry hidden information from the moment they're separated — a little internal label saying “if asked one way, answer up; if asked another, answer down.” The correlations between distant measurements would be no more mysterious than two halves of a torn dollar bill matching up later. The apparent randomness of quantum mechanics would just be our ignorance about these hidden variables. The theory wouldn't be wrong, just incomplete.

It was a beautiful, intuitive escape. For thirty years it sat there as a philosophical preference — local hidden variables versus the quantum formalism — with no way to choose between them.

Bell's inequality

In 1964, John Bell, working at CERN almost as a side project to his day job, did something nobody had thought to do. He took EPR's assumption seriously — that the world is locally real, with hidden variables — and asked what that picture, on its own terms, would predict.

His result is a one-page argument that anyone with a little algebra can follow. Imagine Alice and Bob can each measure their particle along any one of three directions. If the particles carry pre-set answers for every possible measurement (because realism), and if Alice's choice of direction can't affect Bob's outcome (because locality), then the statistical correlations between Alice's and Bob's results, averaged over many runs, must satisfy a particular inequality. Roughly:

The fraction of pairs agreeing at one combination of angles, plus the fraction agreeing at a second, must be at least as large as the fraction agreeing at a third.

It's a bookkeeping fact. It has nothing to do with quantum mechanics. Any theory respecting local realism — every theory of the kind Einstein wanted — has to obey it.

Then Bell computed what quantum mechanics predicts for the same experiment. The answer is plain. Quantum mechanics violates the inequality. The two pictures don't merely differ in interpretation. They disagree in numbers. Experiment can decide.

The experiments

The first tests came from John Clauser and Stuart Freedman in 1972, using pairs of photons emitted by excited calcium atoms. Their detectors clicked in the patterns quantum mechanics predicted; the inequality was violated. But early experiments had loopholes — perhaps the detectors weren't efficient enough, perhaps the two ends could have communicated by some unknown signal during the measurement.

In the early 1980s, Alain Aspect in France closed the most worrying gap: he switched the orientation of his polarizers while the photons were already in flight, too fast for any light-speed signal to coordinate the two sides. Quantum mechanics still won.

Over the next forty years, every remaining loophole was closed in turn. Independent experiments in Delft, Vienna, and Boulder used entangled electrons, photons, and atoms, with random number generators built from cosmic light and even from teenagers playing video games to defeat any hidden coordination. The verdict has held every time. In 2022, Aspect, Clauser, and Anton Zeilinger shared the Nobel Prize for the work.

What the world isn't

Bell's theorem is sometimes described as proving that quantum mechanics is “weird.” That is an understatement. What it proves is that no theory at all, quantum or otherwise, that respects both locality and realism in the EPR sense, can match what we observe. Something has to give.

You can drop locality — accept that distant parts of the universe really are connected in ways that don't fit neatly inside relativity. This is roughly the pilot-wave view, with its non-local guiding field.

You can drop realism — accept that some properties of particles simply do not exist before they are measured. Outcomes are produced in the act of measurement. This is closer to the orthodox Copenhagen reading.

You can drop the assumption that measurement has a single outcome at all, and embrace the many-worlds picture where every possibility happens in a branch of the wavefunction.

You can even drop the assumption of free choice — the idea that experimenters could have picked which measurement to make. This “superdeterminism” loophole is technically open, but most physicists find it less appealing than abandoning realism.

What you cannot do, and this is the part that should stop you, is have the world Einstein wanted. The universe is not a collection of small objects with definite properties, sitting in independent pieces, communicating only through local signals. That picture has been ruled out by experiment.

Einstein died in 1955, nine years before Bell's paper. He went to his grave convinced quantum mechanics could not be the final word. He may yet turn out to be right — the theory remains incomplete in other ways. But on the specific question he raised in 1935, on whether the world could be salvaged as locally real, the answer came back, and it was no. The thought experiment he meant as a reductio became, instead, the cleanest demonstration we have that the universe is stranger than common sense will allow.