Falsifiability
falsifiable · falsification
Karl Popper's criterion for scientific status: a theory is scientific if it forbids something, so that a statable observation would refute it.
In practice
Someone defends a strategy that explains every result. Growth proves it worked, decline proves it was needed sooner. The account rules nothing out, which is what makes it unfalsifiable rather than well supported.
The common mistake
Treating unfalsifiable as a synonym for false. Popper classified such claims as outside science, and thought several were meaningful and some were true.
Falsifiability asks a single question of any theory: what would show this to be wrong? A theory that can name such an observation is falsifiable. One that accommodates every possible outcome is not.
Why the asymmetry matters
No number of white swans establishes that all swans are white. One black swan settles that they are not.
Popper built a whole account of science on this asymmetry. Confirmation is weak because any theory has confirming instances, including bad ones. Refutation is strong because it takes only one. So the useful question about a theory is not what supports it but what it rules out.
The test in practice
A theory passes when it makes a risky prediction — one that could easily have come out the other way, and that a rival theory would not have made. Predicting light bends by a specific measurable amount is risky. Predicting that events will have psychological causes is not, because nothing is excluded.
Popper's charge against the psychoanalysis and Marxist history of his day was this, rather than that they were false: their explanatory success was unlimited, and unlimited explanatory success is a defect.
The limits
Nothing is tested alone. A prediction is derived from the theory plus assumptions about instruments, background conditions and auxiliary hypotheses. When it fails, something in that bundle is wrong, and the experiment does not say which. This is the Duhem–Quine point, and it means a theory can always be rescued by adjusting something else.
Scientists ignore anomalies, correctly. Kuhn argued that a framework discarded at the first failed prediction would never be developed far enough to be useful, and the historical record supports him.
Some good science is not obviously falsifiable, including claims about the distant past and parts of contemporary physics that predict nothing currently measurable.
What survives
Not a sharp line between science and everything else. What survives is the question, which remains the most useful single thing to ask about any claim: what would change your mind? A position whose holder cannot answer that has told you what kind of position it is. See problem of induction and skepticism.
Concept web
Open the full webQuestions
What is falsifiability?
Popper's criterion that a theory is scientific if it forbids something — if a statable observation would refute it. A theory compatible with every outcome fails the test.
Why is falsification stronger than confirmation?
Because of an asymmetry in logic: no number of confirming instances establishes a universal claim, while a single contrary instance refutes it.
What is the main problem with falsifiability?
Nothing is tested alone. A prediction depends on the theory plus assumptions about instruments and background conditions, so a failure does not say which part was wrong.