Settling Einstein's Argument: Bell's Theorem and the Experiments That Tested Reality

In October 2022, the Nobel Prize in Physics went to three experimenters, Alain Aspect, John Clauser, and Anton Zeilinger, for work that resolved an argument Albert Einstein started in 1935 and lost. The argument was about whether the world is, at bottom, as solid and local as it appears, or whether quantum mechanics forces us to give up one of those comforts. For a long time the dispute looked purely philosophical, the kind of thing physicists debated over beer and then set aside to do real work. What makes the story remarkable is that it stopped being philosophy. Someone found a way to put the question to an experiment, and the experiment gave an answer.

Start with the thing that bothered Einstein. Quantum mechanics allows two particles to be prepared in a joint state called entanglement, in which neither particle has a definite property of its own, but the two are correlated. A standard example uses a property of light called polarization, the orientation of a photon’s oscillation. You can create a pair of photons flying off in opposite directions such that their polarizations are linked: whatever result you get measuring one, the other is guaranteed to give the matching result, even though, according to the theory, neither photon had a settled polarization until the measurement happened. Measure one photon here, and the distant partner’s outcome is fixed in that instant, no matter how far away it has traveled.

To Einstein this was intolerable, and in 1935, with Boris Podolsky and Nathan Rosen, he published a paper arguing that it showed quantum mechanics was incomplete. The reasoning ran like this. If I can predict with certainty the result of measuring the far photon by measuring the near one, without disturbing the far photon, then the far photon’s property must have been real all along, sitting there waiting to be found. The theory’s claim that the property did not exist until measured must therefore be missing something. Einstein’s preferred resolution was that the particles carry hidden instructions, set when they were created, that determine in advance how each will respond to any measurement. The apparent spookiness, the instantaneous link across distance that he memorably disliked, would then be an illusion. The correlations would just reflect shared instructions, the way two travelers given sealed envelopes at the start of a trip will find matching messages when they open them, with nothing passing between them at the moment of opening. This view is called local hidden variables: local because nothing travels faster than light, hidden because the instructions are not part of the quantum description.

For nearly thirty years this remained a standoff. Quantum mechanics worked beautifully, but maybe it was an incomplete account sitting on top of a deeper, more sensible, hidden-variable reality. There seemed to be no way to tell the two pictures apart, because they agreed on the everyday predictions. Then in 1964 a physicist at CERN named John Stewart Bell found the crack that let experiment in.

Bell’s insight was to look not at single measurements but at how the correlations behave when the two distant observers are free to choose, independently, among different measurement settings, different angles at which to test the photons’ polarization. He proved a theorem. If the world really runs on local hidden variables, on predetermined instructions with no faster-than-light influence, then the correlations between the two observers’ results, averaged over many runs with various settings, must obey a certain mathematical limit. This limit, now called a Bell inequality, is a ceiling on how strongly the outcomes can be correlated under any local hidden-variable scheme whatsoever. It does not depend on the details of the hidden instructions. Any such theory, however clever, is bound by it.

And here was the payoff: quantum mechanics predicts correlations that break the ceiling. For certain choices of measurement angles, the entangled photons are predicted to be more strongly correlated than any local hidden-variable theory could ever produce. The two views, which had seemed experimentally identical, actually disagreed, and they disagreed about a number you could go out and measure. Bell had turned a metaphysical dispute into a laboratory question. Either the inequality holds, and Einstein’s picture survives, or it is violated, and local hidden variables are ruled out, leaving us with the spooky quantum world.

Doing the experiment was hard, and it took people willing to work on a problem many senior physicists considered a waste of time. John Clauser was among the first. As a young researcher around 1969, he reworked Bell’s idea into a form suited to a real apparatus, and in 1972, with Stuart Freedman at Berkeley, he carried out the first experimental test using entangled photons from calcium atoms. The result violated the inequality, siding with quantum mechanics. It was a striking outcome, but the early experiments had loopholes, gaps that a determined defender of local realism could still slip through. The most important was the locality loophole. If the two measurement stations were close enough, or the settings chosen far enough in advance, then in principle some signal traveling at or below the speed of light could have passed from one side to the other, coordinating the results and faking the quantum correlations without any genuine nonlocality.

Closing that loophole was the achievement most associated with Alain Aspect, working in France in the early 1980s. In his most famous version, completed in 1982, Aspect switched the measurement settings while the photons were already in flight, too late for any light-speed signal from one detector to reach the other before each measurement was done. The choice of what to measure on one side could not have been known in advance by the other side, not even by an influence traveling at the speed of light. The inequality was still violated. The result held up under conditions Einstein’s local picture should not have survived.

Anton Zeilinger and his groups extended this work over the following decades, pushing the experiments to greater distances, testing entanglement between photons kilometers apart, and developing the tools to manipulate entangled states deliberately rather than just observe them. His teams demonstrated entanglement swapping, in which two particles that have never interacted become entangled, and quantum teleportation, in which the quantum state of one particle is transferred to another at a distance. These were not only deeper tests of the foundations but the first instruments of a new field, quantum information science, which treats entanglement as a resource to be engineered.

The remaining loopholes were finally closed together in 2015, when several groups, working independently in the Netherlands, Austria, and the United States, performed loophole-free Bell tests that shut every gap at once. The verdict from all of them was the same as Clauser’s first result in 1972, now beyond reasonable escape. Bell inequalities are violated. Nature does not run on local hidden variables. Einstein’s intuition, that the particles carry predetermined instructions and that nothing spooky crosses the gap between them, is wrong as a description of the world.

It is worth being careful about what this does and does not mean, because the result is easy to overstate. Giving up local hidden variables forces a choice between two cherished assumptions, locality and what philosophers call realism, the idea that properties have definite values before measurement. You can keep one but not both in their classical forms. Different physicists make different peace with this, and the interpretation of quantum mechanics remains genuinely unsettled. What is not in doubt is the experimental fact: the correlations are real, they are stronger than any local realistic theory permits, and they have been confirmed in many laboratories with photons, atoms, and other systems.

One thing entanglement does not allow, despite a great deal of loose talk, is sending messages faster than light. The correlation only shows up when the two observers later compare their records, and that comparison has to travel by ordinary, slower-than-light means. Each observer, looking only at their own side, sees nothing but random results, with no signal in them. The spooky link is real but useless for communication, a subtlety that keeps the whole picture consistent with relativity even as it overturns Einstein’s deeper expectation.

The practical legacy is large and growing. The same entangled correlations that settled the Einstein argument now underpin quantum cryptography, where they let two parties detect any eavesdropper on their key, and they are a core resource in quantum computing, where entanglement between qubits is part of what gives such machines their potential power. The 2022 Nobel citation made the connection explicit, honoring the three not only for testing Bell inequalities but for pioneering quantum information science. A debate that began as Einstein’s complaint about an unfinished theory has become the foundation of a technology.

What I find most satisfying about this history is its shape. A great physicist raised a serious objection. For thirty years it sat as an apparently unanswerable matter of taste. Then one person, Bell, saw how to make it answerable, and a handful of stubborn experimenters spent their careers actually answering it, against the advice of colleagues who thought the question beneath them. The answer turned out to be the more unsettling of the two, the one Einstein hoped to avoid. The world really is, in the precise sense Bell defined, stranger than common sense allows, and we know this not because someone argued it cleverly but because the experiments were done.