In 1967, a chemist named Ray Davis buried 100,000 gallons of dry-cleaning fluid in a gold mine in South Dakota. He was looking for neutrinos from the sun.
Neutrinos are produced in enormous quantities by nuclear fusion in the sun's core. They're nearly massless, have no charge, and pass through almost everything — through rock, through planets, through you right now, in the hundreds of billions per square centimeter per second. The mine shielded Davis's tank from cosmic rays, which would otherwise drown the signal. The fluid could catch neutrinos through a specific reaction: a solar neutrino hits a chlorine-37 atom and converts it to radioactive argon-37. Every few weeks, Davis would bubble helium through the tank and collect the argon atoms that had accumulated. He could count them directly.
Over 25 years, he counted 2,200 of them total.
The problem was that John Bahcall's calculation predicted three times that many. Bahcall was an astrophysicist who had built a detailed model of the sun's interior — the pressure, temperature, and nuclear reaction rates at every depth. His model was the best available account of how the sun works. When Davis got his first results in 1968, the discrepancy was immediate and severe. One-third the predicted flux. The experiment was measuring something, but only a third of what it should be.
This situation — where an experiment contradicts a theory — usually resolves in one of two ways. Either the experiment is wrong (a systematic error, a miscalibrated instrument, something overlooked), or the theory is wrong (a bad assumption, an incorrect input, a flaw in the model). Physicists spent three decades trying both exits. Neither opened.
Davis's experiment was scrutinized relentlessly. He welcomed it. He modified the apparatus, reran the checks, published his methods in exhaustive detail. The number stayed. In the 1990s, helioseismology gave physicists a new way to test Bahcall's solar model: the sun's surface oscillates, and those oscillations encode the internal structure. The oscillation spectrum matched Bahcall's predictions precisely. The sun was exactly as he had modeled it. Both sides of the contradiction were solid.
Bahcall later described those decades as being like "a person who had been sentenced for some heinous crime" — unable to prove his innocence, watching the evidence accumulate against him, not understanding what had gone wrong. He kept recalculating. Davis kept counting. The discrepancy held.
The resolution came from a third possibility that neither man had originally modeled: the neutrinos were changing in transit.
Neutrinos come in three types — electron, muon, and tau — corresponding to the three generations of charged leptons. Davis's experiment could only detect electron neutrinos. Bahcall's model correctly predicted how many electron neutrinos the sun's core produces. What neither accounted for is that neutrinos can spontaneously shift from one type to another as they travel — a quantum process called oscillation. By the time they reached South Dakota, roughly two-thirds had become muon or tau neutrinos that Davis's chlorine detector couldn't see. Davis was measuring exactly the right fraction of exactly the right flux. He was getting one-third because one-third was all there was of the type he could detect.
This was confirmed definitively in 2001 by the Sudbury Neutrino Observatory in Ontario, which used heavy water capable of detecting all three neutrino types. SNO measured the electron neutrino fraction and got Davis's number. Then it measured the total flux — all types combined — and got Bahcall's number. Both were right. The missing two-thirds weren't missing. They had changed form.
The philosophical wrinkle is this: for neutrinos to oscillate, they must have mass. The original Standard Model of particle physics said their mass was exactly zero. This was the deepest implication of the 30-year mystery — not that the experiment or the solar model was wrong, but that one of the foundational assumptions of particle physics needed revision. The deficit was a measurement of something real that nobody had a name for yet. It was pointing at new physics the whole time.
Davis received the Nobel Prize in 2002, at 87, for work he started in his 50s. Bahcall, who died in 2005 before a second prize could recognize the solar model specifically, said when SNO's results were announced that he felt like dancing. Arthur McDonald and Takaaki Kajita shared the 2015 Nobel for the oscillation discovery itself.
What stays with me isn't the resolution. It's the 30 years in the middle. A measurement that kept coming back wrong. A theory that kept checking out. Both sides held. The two people most responsible for the contradiction — Davis at his tank, Bahcall at his equations — maintained confidence in their results without being able to explain the gap between them. The discrepancy wasn't a sign that one of them had made an error. It was a message, arriving in fragments, from a phenomenon that didn't have a name yet.
The experiment was incomplete not because it was poorly designed but because it could only see one-third of what was being sent. The question is whether that's a special circumstance or a general condition.