In 1919, Karl Popper noticed something that bothered him about several popular theories of the time. Freudian psychoanalysis, Adlerian psychology, and Marxist historical theory all shared a curious property: they could explain anything. Whatever happened, the theories could accommodate it. A patient improved, the theory explained it. A patient got worse, the theory explained that too. Any historical event could be reinterpreted as confirming the Marxist framework. The theories seemed powerful precisely because nothing could refute them.
Popper contrasted this with Einstein’s general relativity, which made a very specific prediction: light would bend around the Sun by a precise amount during a solar eclipse. If the 1919 eclipse observations had shown a different amount, general relativity would have been falsified. The theory stuck its neck out. It was vulnerable to being wrong, and it survived the test.
This contrast led Popper to his central insight: what distinguishes science from non-science is not whether a theory can be verified, but whether it can be falsified. The criterion he developed (falsifiability) became one of the most influential ideas in the philosophy of science.
The Core Principle
Falsifiability, as Popper defined it, is the property of being testable in a way that could show the claim to be false. A falsifiable statement is one for which there exists at least one possible observation or experimental outcome that would, if it occurred, count as evidence against the statement.
This is a logical property, not a practical one. A claim can be falsifiable even if it is true and will never actually be falsified. “All ravens are black” is falsifiable because a single white raven would count against it. The claim has been tested repeatedly and has survived, making it well-supported, but it remains falsifiable in principle.
Contrast this with “There exists a dragon in this room that is invisible, intangible, and undetectable by any instrument.” No observation could count against this claim. It is not falsifiable, and therefore, by Popper’s criterion, it is not scientific.
The principle extends to compound systems. A claim that can be made unfalsifiable by adding auxiliary hypotheses (“the test failed, but that’s because we didn’t account for variable X”) can be protected from falsification indefinitely. Popper called this the problem of immunization: wrapping a hypothesis in enough escape clauses to make it untestable in practice.
Why Verification Isn’t Enough
Before Popper, many philosophers of science emphasized verification: a good scientific claim is one supported by positive evidence. The more observations that confirm a claim, the better it is established.
Popper rejected this as the demarcation criterion. His problem, formulated as the problem of induction (originating with David Hume), is that no number of confirming observations can logically prove a universal claim. A million observed black ravens do not prove that all ravens are black; the next one might be white. But a single white raven can definitively show that “all ravens are black” is false.
Popper argued that science advances not by accumulating confirming observations but by eliminating false theories through disconfirmation. A theory that has survived many attempts at falsification (and not just any attempts, but severe tests designed to challenge it) is corroborated, even though it is never conclusively proven. Science does not achieve certainty; it achieves progressively better-tested approximations to truth.
The Asymmetry of Testing
Popper’s insight relied on a logical asymmetry in how general claims relate to observations. Consider the statement: “All copper conducts electricity.”
– To verify it completely, you would need to test every piece of copper everywhere in the universe, impossible. – To falsify it, you need only find one piece of copper that does not conduct electricity.
This asymmetry means that falsification provides a logically decisive result that verification cannot. One counterexample defeats the universal claim. Any number of confirming examples cannot establish it with certainty.
This is why Popper elevated falsifiability over verifiability: falsification has logical teeth that verification lacks. A good scientific theory is one that generates specific predictions that observation could in principle overturn, and the theory gets stronger each time a severe test fails to overturn it.
Popper’s Criterion in Practice
Applying Popper’s criterion to real science reveals both its power and its complications. Most working scientists would accept the following as scientific:
Germ theory predicts that specific microorganisms cause specific diseases. The prediction can be tested: isolate the organism, introduce it into a healthy host, observe whether disease results. The theory has been tested thousands of times and has passed.
The theory of natural selection predicts that populations under selection pressure will change over generations in the direction that improves reproductive fitness. This can be tested in the laboratory (Lenski’s long-term E. coli evolution experiment, running since 1988) and observed in nature. The prediction has been confirmed repeatedly.
General relativity predicted gravitational waves — ripples in spacetime produced by accelerating masses. The prediction was specific enough that a non-detection after sufficiently sensitive instruments were built would have been evidence against the theory. LIGO’s detection in 2015 confirmed the prediction a century after it was made.
In each case, the theory made specific, potentially falsifiable predictions that subsequent observation tested.
Criticisms and Complications
Popper’s falsifiability criterion has been criticized, refined, and partially replaced by philosophers of science, while remaining influential in practice.
The Duhem-Quine problem: In practice, scientific theories are never tested in isolation. A prediction depends on the core theory plus a large number of auxiliary assumptions (about instruments, background conditions, and other variables). When a test fails, you cannot be certain whether the core theory is wrong or one of the auxiliary assumptions is false. This makes clean falsification logically complicated.
Pierre Duhem, a physicist and philosopher writing before Popper, had already noted that a hypothesis “is never isolated from a whole set of hypotheses.” When an observation contradicts a prediction, you can choose to reject any part of the system, not necessarily the main hypothesis. Scientists routinely protect core theories from disconfirmation by adjusting auxiliary assumptions, and often correctly so.
Thomas Kuhn’s challenge: Thomas Kuhn, in his 1962 book The Structure of Scientific Revolutions, argued that science does not proceed by continuous falsification but by paradigm shifts. Normal science works within an accepted framework (paradigm) and treats anomalous observations as puzzles to be solved within the paradigm, not as falsifications of it. Major theoretical changes happen rarely and involve sociological and psychological factors beyond pure logic. Kuhn’s account is more descriptively accurate than Popper’s, but it arguably makes science seem less rational.
Imre Lakatos’s research programmes: Lakatos attempted to reconcile Popper and Kuhn by proposing that scientists work within “research programmes” (clusters of theories with a protected “hard core” of core assumptions and a flexible “protective belt” of auxiliary hypotheses). A programme is “progressive” if it generates novel confirmed predictions and “degenerative” if it only accommodates observations after the fact. This framework preserves Popper’s emphasis on predictive power while acknowledging Kuhn’s observation that scientists do not abandon core theories at the first anomaly.
Unfalsifiable but scientific? Some areas of physics have faced criticism for being unfalsifiable. String theory and many versions of the multiverse hypothesis are difficult or impossible to test with current or foreseeable technology. Critics invoke Popper to argue these are not science. Defenders argue that mathematical consistency, explanatory power, and indirect testability are sufficient to count as scientific activity, or that testability is a matter of degree and current inaccessibility does not permanently remove a claim from science.
What Falsifiability Does Not Mean
Falsifiability is a criterion for what can be scientific, not a criterion for what is true. An unfalsifiable claim can be true. Ethical claims, mathematical propositions, and metaphysical assertions can all be true without being falsifiable. They simply operate by different standards of justification.
Falsifiability also does not mean that falsified theories are worthless. Newton’s mechanics were falsified in the domain of very high velocities and very large masses, replaced by special relativity and general relativity. But Newtonian mechanics are still used constantly in engineering, physics education, and everyday calculation, because they are excellent approximations within their domain. A theory’s falsification in one domain does not erase its accuracy in another.
What is falsifiability in simple terms?
Falsifiability means that a claim can be tested in a way that could show it to be wrong. A falsifiable claim makes specific predictions, and if those predictions turn out to be false, the claim is falsified. For example, u0022all swans are whiteu0022 is falsifiable — one black swan disproves it. u0022Invisible forces control everythingu0022 is not falsifiable because no observation could count against it.
Who invented the concept of falsifiability?
Karl Popper (1902–1994), an Austrian-British philosopher of science, developed the concept of falsifiability in the 1930s, most fully in his 1934 book u003cemu003eLogik der Forschungu003c/emu003e (published in English as u003cemu003eThe Logic of Scientific Discoveryu003c/emu003e in 1959). He proposed it as the criterion for distinguishing science from non-science, what he called the demarcation problem.
Is a non-falsifiable theory always wrong?
No. Unfalsifiable claims can be true; they are simply outside the domain of empirical science. Mathematical theorems, ethical claims, and religious beliefs may all be non-falsifiable in the Popperian sense without being false. Falsifiability is a criterion for what counts as empirical science, not a criterion for what counts as truth.
What is the Duhem-Quine problem?
The Duhem-Quine problem is the observation that scientific tests never test a single hypothesis in isolation; they test a hypothesis together with many auxiliary assumptions. When a prediction fails, it is not automatically clear which part of the overall system is wrong: the main hypothesis, or one of the auxiliary assumptions. This means that in practice, falsification is not as clean or decisive as Popper’s criterion suggests.
Is string theory scientific if it can’t be tested?
This is actively debated among physicists and philosophers of science. By Popper’s strict criterion, aspects of string theory that make no testable predictions would not count as science. Defenders argue that string theory’s mathematical coherence and its ability to reproduce known physics indirectly constitute scientific activity. Most practicing scientists take a more pragmatic view: untestable theories are scientifically less valuable, but mathematical exploration is part of the scientific enterprise even when direct testing is not yet possible.
How is falsifiability different from skepticism?
Falsifiability is a criterion for what counts as a scientific claim; it is about the logical structure of theories. Skepticism is an epistemic attitude, a disposition to require evidence before accepting claims. They are related but distinct. A committed falsificationist emphasizes that the best scientific theories are those that have survived the most severe attempts to falsify them. A skeptic emphasizes demanding evidence for any claim. Both value evidence and testability, but from different angles.
Sources
Popper, K.R. (1959). The Logic of Scientific Discovery. Hutchinson. (Original German edition 1934.)
Kuhn, T.S. (1962). The Structure of Scientific Revolutions. University of Chicago Press.
Lakatos, I. (1978). The Methodology of Scientific Research Programmes. Cambridge University Press.
Duhem, P. (1954). The Aim and Structure of Physical Theory. Princeton University Press. (Original French edition 1906.)
Quine, W.V.O. (1951). Two Dogmas of Empiricism. The Philosophical Review, 60(1), 20–43. doi:10.2307/2181906
Pigliucci, M., & Boudry, M. (Eds.) (2013). Philosophy of Pseudoscience: Reconsidering the Demarcation Problem. University of Chicago Press.
This article is part of our framework exploring Knowledge — how science, mathematics, and language help us understand reality.