Falsifiability

Falsifiability is the property of a claim, hypothesis, or theory that some possible observation or experimental result would contradict it. In philosophy of science, it is especially associated with Karl Popper, who proposed it as a criterion for distinguishing empirical science from non-scientific systems of thought. A falsifiable theory need not be false: it must exclude some possible observable outcomes and therefore expose itself to tests that it could fail. Falsifiability is a capacity for empirical refutation, whereas falsification is the actual rejection of a claim on the basis of evidence and accepted testing assumptions. (plato.stanford.edu)

Logical basis

Falsifiability rests on an asymmetry between establishing a universal generalization and finding a counterexample to it. However many white swans have been observed, those observations do not logically establish that every swan is white. An accepted observation of a non-white swan, by contrast, contradicts the unrestricted claim that all swans are white. The example concerns the logical relationship between statements; whether a reported bird really is a swan, and whether its colour has been reliably observed, are further empirical questions. (iep.utm.edu)

The underlying form of deductive reasoning is modus tollens:

Here $T$ is a theory and $P$ an observational consequence. If the theory entails the prediction and the prediction is false, the theory cannot be true. Observing $P$, however, does not deductively establish $T$, because another theory might entail the same prediction. This distinguishes deductive refutation from inductive reasoning based on accumulated positive instances. (iep.utm.edu)

A merely imaginable contradiction is insufficient: the potential counterexample must be expressible as an empirical observation statement. Popper called relevant observation reports basic statements and described statements incompatible with a theory as its potential falsifiers. Their acceptance is itself open to criticism rather than resting on infallible observation. (plato.stanford.edu)

Historical development

Popper developed falsifiability against the background of the problem of induction associated with David Hume. His approach rejected the idea that accumulating observations could conclusively justify universal scientific laws. Instead, scientific inquiry would propose conjectures and subject their consequences to critical testing. (plato.stanford.edu)

In his retrospective account, Popper contrasted theories that appeared capable of accommodating almost any outcome with Albert Einstein’s gravitational theory, which exposed itself to possible failure through definite observational predictions. His concern was not simply whether a theory had supporting examples, but whether its supporters could identify evidence that would count against it. (tildesites.bowdoin.edu)

This project differed from the verification-oriented approaches associated with logical empiricism. Falsifiability was intended as a criterion of scientific status, not a general criterion of meaningful language. Popper did not equate non-scientific ideas with nonsense: metaphysical ideas could be meaningful and could contribute to the development of science without themselves being empirically falsifiable. (iep.utm.edu)

Falsifiability and scientific method

Falsifiability is a logical property; falsificationism is a methodological position about how scientific inquiry should proceed. In Popper’s account, scientists formulate theories, derive consequences, and seek demanding tests rather than merely collect favourable instances. Predictions are especially informative when their failure would create a serious difficulty for the theory. (tildesites.bowdoin.edu)

Popper described successful survival of testing as corroboration. Corroboration records how a theory has performed under critical examination; it is not conclusive verification or, in his account, a probability that the theory is true. A well-corroborated theory remains open to revision or replacement. (plato.stanford.edu)

The empirical content of a theory is connected to what it rules out. A precise prediction can exclude more outcomes than a vague statement that something will happen eventually. Popper therefore recognized degrees of testability: theories can expose themselves to stronger or weaker possible refutations. Adjustments that merely explain away every adverse result can reduce that exposure. (tildesites.bowdoin.edu)

Auxiliary assumptions and fallible observations

Scientific predictions usually depend on more than the theory being tested. They also require assumptions about initial conditions, instruments, experimental arrangements, and other background knowledge. The inference is therefore more accurately represented as

where $A$ represents auxiliary assumptions and $I$ initial conditions. If $P$ fails, the deductive conclusion is

not necessarily $\neg T$. Something in the conjunction is mistaken, but logic alone does not identify which component. (mpra.ub.uni-muenchen.de)

This difficulty is central to the Duhem–Quine thesis and related discussions of testing holism. An apparent counterexample may expose a faulty instrument, an inaccurate background assumption, or a defect in the principal theory. Investigating these alternatives is not automatically an evasion of testing. The methodological problem is distinguishing independently defensible corrections from changes introduced solely to protect a theory. (mpra.ub.uni-muenchen.de)

Consequently, falsification in scientific practice is not an infallible mechanical procedure. Popper acknowledged the fallibility of observation reports and did not treat every isolated discrepancy as sufficient grounds for abandoning a theory. Acceptance of an empirical refutation depends on critical examination of the evidence and testing conditions. (plato.stanford.edu)

Probabilistic hypotheses

The relation between falsifiability and probability is more complicated than the universal-swan example suggests. As an illustration, a model of independent fair coin tosses assigns positive probability to every finite heads-and-tails sequence. Even a very unusual sequence therefore need not logically contradict that model. Evidence can nevertheless change the relative credibility of competing probabilistic hypotheses. (plato.stanford.edu)

In statistical hypothesis testing, results can indicate incompatibility between data and a specified model, but statistical rejection is not deductive proof of falsehood. A p-value does not give the probability that the null hypothesis is true, and scientific conclusions cannot be determined solely by whether it crosses a numerical threshold. (amstat.org)

Bayesian inference offers a different account of evidential assessment. It updates the probabilities assigned to hypotheses in light of evidence and background assumptions. Bayesian confirmation theory can represent both increases and decreases in support without requiring each observation to produce a binary verdict of verification or falsification. This contrasts with Popper’s rejection of interpreting corroboration as the probability of a theory’s truth. (plato.stanford.edu)

Demarcation and criticism

The demarcation problem concerns how science should be distinguished from non-science, including pseudoscience. Falsifiability is an influential proposed criterion, but it is not generally accepted as a complete solution. Critics note that some pseudoscientific claims make testable predictions, while scientific work often continues despite unresolved anomalies. The logical form of an isolated claim may therefore be insufficient to assess a broader practice of inquiry. (plato.stanford.edu)

Thomas Kuhn emphasized that much ordinary scientific work consists of solving problems within an accepted framework rather than repeatedly trying to overthrow it. From this perspective, tolerating an anomaly can be part of productive research rather than evidence of an unscientific attitude. (plato.stanford.edu)

Imre Lakatos shifted assessment toward sequences of theories organized into research programmes. A programme is progressive when theoretical changes produce additional empirical predictions and at least some receive support. It is degenerating when changes primarily accommodate difficulties without corresponding predictive progress. His account made the history of a programme, not merely an individual theory’s logical falsifiability, central to evaluation. (plato.stanford.edu)

These debates distinguish several questions that are often conflated: whether a claim could conflict with evidence, whether it has actually survived demanding tests, whether its proponents respond critically to contrary results, and whether the surrounding research programme generates empirical progress. Falsifiability directly addresses the first question; the others require additional methodological and historical judgments. (plato.stanford.edu)

References

1. Karl Popperplato.stanford.edu
2. Popper, Karl: Philosophy of Scienceiep.utm.edu
3. Karl Popper "Science as Falsification" (1963)tildesites.bowdoin.edu
4. Falsificationismmpra.ub.uni-muenchen.de
5. Confirmationplato.stanford.edu
6. American Statistical Association Releases Statement on Statistical Significance and P-Valuesamstat.org
7. Science and Pseudo-Scienceplato.stanford.edu
8. Imre Lakatosplato.stanford.edu