The man who made it testable
In 1964, an Irish physicist named John Bell did something nobody thought possible: he found a way to *measure* the difference between Einstein's worldview and quantum mechanics. For thirty years the EPR debate had felt like arguing about taste. Bell saw that if particles really carried hidden, pre-set answers — local hidden variables — then the statistics of many measurements would obey a strict limit. Quantum mechanics predicted those same statistics could break that limit. One number, two predictions. Nature would decide. The limit he derived is now called Bell's inequality.
The brilliance is that Bell's limit follows from local realism alone, with no reference to quantum mechanics at all. Any universe built from objects with definite local properties — no matter what the hidden details are — must obey it. So if an experiment beats the limit, no clever local-hidden-variable model can be patched in to save common sense. That is what makes it a genuine test of reality, not of one particular theory.
The trick: measure at different angles
Here is the clever twist. In the simple coin story, Alice and Bob always measured the "same question." Bell's idea was to let each of them randomly choose which question to ask — for example, by orienting a measuring device at one of a few different angles. Think of the apparatus like the magnet in a Stern–Gerlach experiment, which sorts particles into up or down along whatever direction you point it. Alice picks an angle, Bob picks an angle, and they record their two outcomes. Repeat thousands of times with fresh entangled pairs, then study how often their answers agree as the angle settings change.
Why does varying the angle matter so much? Because if each particle secretly carries a pre-written answer for *every* possible angle Alice or Bob might choose, those pre-written answer sheets can only be coordinated so well. There is a ceiling on how strongly the agreements can swing as the angles change — that ceiling is Bell's inequality. Quantum particles, which have no answer sheets and decide on the spot, can swing harder. The mismatch is exactly what an experiment looks for.
The number nature must respect — or not
The most-used practical version is the CHSH inequality, named for four physicists who recast Bell's idea for real labs. You combine the agreement statistics from four particular angle pairings into a single score, usually written S. The rule from local realism is stark and simple:
Local hidden variables (Einstein's world): |S| <= 2 Quantum mechanics predicts (best angles): |S| = 2 * sqrt(2) (about 2.83) Measure S in the lab. If S stays under 2 -> Einstein could be right. If S exceeds 2 -> no local-hidden-variable story can fit.
Read that again, because it is the whole ballgame. If the world runs on local hidden variables, the score can never climb above 2 — that is a mathematical certainty, true of any conceivable such theory. Quantum mechanics says that with the right angle choices the score climbs to about 2.83, comfortably past the wall. So the experiment is not subtle: just measure S and see which side of 2 it lands on. The stronger-than-classical agreements that push it past 2 are the signature quantum correlations of entanglement.
From thought experiment to lab bench
Bell handed physics a gift: a philosophical quarrel rewritten as a measurement anyone with the right equipment could perform. A run that probes this score is now called a Bell test. Notice what is and isn't being tested. A Bell test does not assume quantum mechanics is right; it pits local realism against whatever nature actually does. If S exceeds 2, then at least one of EPR's two cherished assumptions — locality, or realism — must be false. Common sense itself goes on trial. All that remained, after 1964, was for someone to build the apparatus and run it. The next guide is the story of who did, and what they found.