What is falsifiability? Popper's test for a scientific claim

Falsifiability is the property a claim has when some possible observation could show it to be false. "All swans are white" is falsifiable, because a single black swan would refute it. The philosopher Karl Popper argued that this is what separates scientific claims from others: a scientific theory sticks its neck out, saying in advance what the world must not do, and a theory that fits every possible result tells us nothing.

In short

• A claim is falsifiable if some observation, at least in principle, would contradict it. Falsifiable doesn't mean false.
• Popper saw an asymmetry: no number of confirming cases proves a general law, but one solid counterexample can refute it.
• The idea is still the most quoted test for science, but philosophers agree it is too simple on its own, because a failed prediction can be blamed on something other than the theory.

What does it mean for a claim to be falsifiable?

A claim is falsifiable when it rules something out. It needs only one conceivable observation that would contradict it, not an actual one. "Water at sea level boils at 100 °C" is falsifiable: a pot of pure water boiling at 90 °C at sea level would refute it. That it has never happened doesn't matter. A falsifiable claim can be true, and the best scientific claims are both falsifiable and, so far, unrefuted.

The idea rests on an asymmetry in logic. Seeing a thousand white swans doesn't prove that all swans are white; the thousand-and-first might be black. But one black swan does prove that not all swans are white. Europeans who had only ever seen white swans learned this when Dutch explorers reported black swans in Australia in 1697. Generalizing from cases is induction, and it never reaches certainty, a problem the philosopher David Hume set out in the 18th century. Popper's answer, in the Stanford Encyclopedia of Philosophy's words, was that a universal law can't be verified by experience, but a genuine counter-instance can refute it.

A quick way to see the difference is to ask of any claim: what would show this to be wrong?

Claim	Falsifiable?	What would refute it
All copper conducts electricity	Yes	A piece of copper that doesn't
This drug lowers blood pressure more than a placebo	Yes	A fair trial in which it doesn't
Everything happens for a reason	No	Nothing: any event can be given a reason afterwards
You will meet someone important soon	Barely	"Soon" and "important" are too loose for any result to count against it
There is a teapot orbiting the Sun too small for any telescope	No, in practice	Nothing anyone could observe

An unfalsifiable claim isn't necessarily false or meaningless. It may be a matter for philosophy, ethics or faith. Popper's point was narrower: it isn't a scientific claim, because no test could ever tell against it.

Why did Popper care so much?

In Vienna in 1919, the young Popper was struck by a contrast. Einstein's theory of gravity made a risky prediction that observation could have contradicted. The psychoanalytic theories of Sigmund Freud and Alfred Adler, by contrast, seemed able to explain any behavior at all: a man who saves a drowning child and a man who pushes one in could both be accounted for. Their followers saw confirmations everywhere. Popper concluded that this was a weakness, not a strength. He set out the criterion in Logik der Forschung (1935), translated as The Logic of Scientific Discovery in 1959, and returned to it in Conjectures and Refutations (1963).

He called the question of where science ends the problem of demarcation. His answer was that a theory is scientific if it can clash with observation, and that scientists should try hard to refute their theories rather than collect support for them. That makes science a sequence of bold guesses and serious attempts to knock them down, with the survivors kept for now. It also means any working hypothesis should be written so that a test can fail it.

How did the 1919 eclipse put relativity at risk?

General relativity, published by Albert Einstein in 1915, predicted that starlight passing close to the Sun would be bent by its gravity, by about 1.75 seconds of arc for a ray grazing the Sun's edge. A Newtonian calculation gave half that. Stars that close to the Sun can be photographed only during a total solar eclipse, when the Sun's glare is blocked.

The test, step by step:

1. Fix the prediction before looking. Relativity says 1.75 arcseconds; Newton's theory, about half; no bending at all, zero. Any of the three results was possible.
2. Photograph the same stars twice. Once at night, months earlier, when the Sun is elsewhere in the sky, and once during the eclipse of 29 May 1919, with the Sun in front of them. British teams went to Sobral in Brazil and to the island of Príncipe off West Africa.
3. Measure the shift. Compare each star's position on the two sets of plates and work out how far the eclipse images are pushed outward from the Sun.
4. Compare. The best Sobral plates gave about 1.98 arcseconds and the Príncipe plates about 1.61, both far closer to Einstein's figure than to Newton's. The results were announced in London on 6 November 1919 (Wikipedia).

The point for falsifiability is step 1. Had the plates shown zero or the Newtonian value, relativity would have been in serious trouble, and Einstein had said in advance what result he expected. The measurements were rough: cloud at Príncipe and a distorted telescope at Sobral, whose plates were set aside, led historians later to question the analysis. But a 1979 remeasurement of the Sobral plates supported the original result, and radio measurements since the late 1960s have confirmed the full bending far more precisely.

Is astrology falsifiable?

Popper used astrology as a standard example of a pseudoscience, and the Stanford Encyclopedia of Philosophy treats it the same way. His complaint was not that astrology can't be tested. Astrologers, he wrote, made their predictions vague enough to survive anything, and explained away the ones that failed.

When astrology's claims are made precise, they can be tested, and they have been. In a double-blind study published in Nature in 1985, the physicist Shawn Carlson asked astrologers to match people's birth charts to their personality profiles. They did no better than chance (Nature). So the sharper lesson is that a precise version of a claim can be falsified, and what makes a practice unscientific on Popper's view is refusing to accept the refutation.

What are the problems with falsifiability?

No theory is tested alone. A prediction always depends on extra assumptions, called auxiliary hypotheses: that the telescope works, that no unseen body is interfering, that the sample is typical. When a prediction fails, logic says only that something in the bundle is wrong, not which part. This is the Duhem–Quine problem, named after the physicist Pierre Duhem and the philosopher W. V. O. Quine.

Planetary history shows both sides. In the 1840s, Uranus drifted from the orbit Newton's laws predicted. Instead of rejecting Newton, astronomers assumed an unseen planet, and in 1846 Neptune was found close to where Urbain Le Verrier calculated it should be. Le Verrier then explained a similar drift in Mercury's orbit by an unseen planet, named Vulcan, which was never found. Mercury's orbit was explained only in 1915, by relativity. Protecting a theory by blaming an assumption was right the first time and wrong the second, and nothing in falsification alone says which is which.

Still argued

How much of Popper's criterion survives is debated. Thomas Kuhn argued in The Structure of Scientific Revolutions (1962) that working scientists don't drop a theory at the first anomaly; they keep it until a better one is available, and that this is reasonable. Imre Lakatos argued that theories are protected by a "protective belt" of auxiliary assumptions, and that the real test is whether a research program keeps predicting new facts or only patches old failures. Defenders of Popper reply that he himself warned against rescuing theories by ad hoc changes, and that falsifiability remains the clearest rule of thumb for spotting a claim that can't lose. Most philosophers now see it as one important mark of science among several, not a complete definition.

How can you use falsifiability day to day?

The practical form of Popper's idea is a single question: what would change your mind? A claim whose supporters can name a result that would count against it is open to evidence. One whose supporters can't, or who explain every failure away after the fact, is protected from it. Watch for claims that are made vaguer each time they miss.

The same habit works on one's own beliefs, where it runs against the pull to notice only what fits, which is confirmation bias. Falsifiability also pairs naturally with Occam's razor: among explanations that survive the tests, prefer the one with fewer extra assumptions. And science as a whole works best when its claims are falsifiable and its practitioners really try to falsify them.

Check yourself

• Without looking back, explain why a thousand white swans prove less than one black swan.
• In the 1919 eclipse test, which step made relativity falsifiable, and what result would have counted against it?
• Why did saving Newton's laws by positing a new planet work for Uranus but fail for Mercury?

Sources

• Karl Popper — Stanford Encyclopedia of Philosophy
• Science and Pseudo-Science — Stanford Encyclopedia of Philosophy
• Eddington experiment — Wikipedia
• A double-blind test of astrology — Shawn Carlson, Nature, 1985
• Duhem–Quine thesis — Wikipedia
• Falsifiability — Wikipedia