Falsification

In 1919, a British expedition photographed a solar eclipse to see whether starlight bent as it passed the sun. Einstein's general relativity said it would, by a specific amount. If the plates had shown otherwise, the theory would have been finished. A seventeen-year-old in Vienna noticed this and never stopped thinking about it.

What made the contrast vivid for Karl Popper was what else was going on around him. Vienna in 1919 was full of Marxists explaining the political situation, Adlerians explaining behaviour by inferiority, and Freudians explaining everything. Popper had himself been a Marxist briefly, and was present when police shot young socialists at a demonstration: an event the theory absorbed smoothly as an inevitable episode in the class struggle. Its ability to absorb it was, he came to think, the problem.

His example is deliberately plain. A man pushes a child into a river to drown her. Another man dives in and risks his life to save a child. Freudian theory explains the first by repression and the second by sublimation, and it could have explained the reverse with equal ease. Adlerian theory explains the first by the man proving he dares and the second by the man proving he dares. Every conceivable human action confirms them. Popper's conclusion is that this apparent strength is a fatal weakness: a theory compatible with every possible observation forbids nothing, and a theory that forbids nothing says nothing about the world.

So the criterion: a theory is scientific to the extent that it is falsifiable. To the extent that there is some observation, possible in principle, that would refute it. The more a theory forbids, the better it is, because it is saying more and risking more. Confirmations are worthless when they were easy to obtain; they count only when they were risky predictions that could have failed.

Two things about this are constantly misreported. First, it is a criterion of demarcation, not of truth or of value. Popper is separating science from non-science, not sense from nonsense. He says explicitly that unfalsifiable ideas may be true, may be valuable, and may later become testable; atomism was metaphysics before it was physics. Calling something unfalsifiable is not calling it false. Second, he is not a naive falsificationist. He knew perfectly well that any test relies on auxiliary assumptions (the instrument is calibrated, the sample is clean, the background conditions hold) so a failed prediction never tells you by itself which part to blame. That is the Duhem-Quine problem and it is the hardest objection to his position. His answer is methodological rather than logical: scientists should decide in advance what they will accept as a refutation, and should not rescue a theory by bolting on modifications whose only motivation is the rescue.

The whole structure also solves, or dissolves, an old problem. Hume had shown that no number of observed instances can logically establish a universal law: a million white swans do not entail that all swans are white. Every attempt to justify induction had either assumed what it was proving or appealed to induction again. Popper simply accepts Hume's negative conclusion in full and argues that science never depended on induction in the first place. Scientists do not accumulate instances and generalise. They guess boldly, then try to kill the guess. What survives severe testing is not confirmed; it is corroborated, which reports only that it has not yet failed. He is insistent on the vocabulary, because 'confirmed' smuggles back the inference he is refusing.

The same shape reappears in his politics. The traditional question is who should rule (the wise, the workers, the nation) and any answer invites entrenching whoever qualifies. Popper substitutes a different question: how can institutions be arranged so that bad rulers can be removed without violence? Democracy then stops being a claim about the wisdom of majorities and becomes an error-correction mechanism, valuable for the same reason falsifiability is: it lets you find out you were wrong at a bearable cost.

The most portable thing here is not the philosophy of science. It is a pattern-recognition skill. When someone tells you their system works for everyone, and that every apparent failure is explained by the person having done it wrong, they have not made a strong claim. They have made an empty one, and no amount of looking at the world will now tell you whether to believe it.