Falsification

Falsification is the demonstration that a statement is invalid. Logically it is decisive in a way that confirmation never is. One contrary case settles a general claim.

Verification is the demonstration that a statement is true. For general statements it is unreachable, which is the asymmetry the whole method rests on. Confirming instances accumulate without ever completing.

A falsifier is a statement or an experimental result that would establish a claim as false. Naming it in advance is what makes a claim testable. A claim whose falsifier cannot be named is not testable.

Falsificationism is Popper's position that science proceeds by refutation rather than by proof. It replaces the accumulation of support with the survival of attack. It is a rule about method rather than about truth.

Karl Popper made falsification the central element of testing theories. His aim was to separate science from claims that explain everything. The criterion is simple to state and is evaded constantly.

Degree of corroboration measures how severely a theory has been tested without failing. It is explicitly not a probability of truth. A theory tested weakly and never refuted is not corroborated.

A basic statement describes a concrete observation and serves as the logical point at which a theory meets the world. Everything turns on whether such statements are independent of the theory. They are not, which is treated below.

An auxiliary hypothesis is an assumption added alongside the main claim. Adding one after a failed test can save a theory from refutation. Whether that is legitimate is the central practical question.

The null hypothesis is the assumption of no effect, against which observed data are tested. It is the statistical form of the falsification idea. Rejecting it is the commonest scientific act there is.

Internal validity is the property of a study that its result is not produced by something other than the factor studied. Without it a refutation refutes the design. It decides whether a failed test means anything.

Refutation is the demonstration of falsity through a contrary case. It is the outcome falsification aims at. It is also what every party with a stake in the theory is motivated to avoid.

A universal statement claims something for all cases without limit. It cannot be verified because the cases do not run out. Scientific laws take this form, which is why the asymmetry matters.

Critical realism is the philosophical position that there is a world independent of our theories, which our theories can get wrong. It is the ground on which refutation makes sense. Without it, failure would only mean incoherence.

It presents refutation as a clean logical operation. It leaves out the logic that makes it work, the reasons a failed test never identifies what failed, the whole repertoire of evasion, the statistical reality of testing, and the contemporary evidence that refutation largely does not happen.

Modus tollens is the inference from a false consequence to a false premise. It is the logical form behind every refutation. Introductory accounts assert the conclusion without naming the form.

The Duhem-Quine thesis holds that a hypothesis is never tested alone but always with a network of assumptions. Accounts treat hypotheses in isolation. When the test fails, logic says something in the network is false and does not say what.

Falsifiability is the property that some observation could refute a claim. Texts focus on the act of falsifying rather than on the property. The property can be assessed before any experiment and is the more useful of the two.

An ad hoc hypothesis is an assumption added purely to rescue a threatened theory. Adding it restores agreement with the data and explains nothing new. It is the most common form of evasion and is not treated.

Theory-ladenness is the fact that observation statements carry theoretical assumptions. Basic statements are treated as pure facts. If the observation already assumes the theory, the test is not independent of it.

An immunisation strategy defends a theory against refutation by redefinition or by added assumptions. Accounts do not treat it. It is what actually happens when a valued theory meets contrary data.

Imre Lakatos developed naive falsificationism into an account of research programmes, in which single refutations do not decide. The development is absent from the accounts. It is the response to the fact that scientists do not abandon theories on one failure.

A research programme is a sequence of theories sharing a hard core, judged over time rather than by single tests. Focus rests on isolated hypotheses. Judging a programme asks whether it keeps predicting new things.

The protective belt is the layer of adjustable auxiliary assumptions around an untouchable core. The distinction between core and belt is absent. It explains which modifications are legitimate and which are evasion.

The problem of induction is the reason Popper replaced verification with refutation. His epistemological motive is usually omitted. Without it, falsificationism looks like a preference rather than a solution.

Underdetermination is the situation where the same data support incompatible theories. It is not treated. Where it holds, a refutation of one theory does not select among the rest.

Quine argued that beliefs face evidence as a whole system, so any statement can be held true if enough else is adjusted. He is the main proponent of the holism behind the Duhem-Quine thesis. He is not named.

Pierre Duhem showed from the history of physics that experiments never test a single hypothesis. He is the historical origin of the auxiliary-hypothesis problem. The problem appears without its source.

Abduction is inference to the best explanation, alongside induction and deduction. It is how hypotheses are generated in the first place. Accounts of testing omit where the tested thing comes from.

Paul Feyerabend attacked the idea that any fixed methodological rule, falsification included, describes successful science. He is the strongest available objection. He is not presented as one.

Bayesian inference updates degrees of belief rather than deciding between refuted and not refuted. It is the probabilistic alternative to the binary account. It is absent.

A prior probability is what was believed before the test. It is excluded from the logical account. It explains why the same failed test convinces one researcher and not another.

A type I error is the mistaken rejection of a true null hypothesis. It is a refutation that is itself false. The theoretical account has no place for it.

A type II error is retaining a false assumption because the test was not powerful enough to detect the effect. It is a refutation that failed to occur. Most non-refutations in practice are of this kind.

Thomas Kuhn described science as changing through paradigms rather than through accumulating refutations. His account is sociological where Popper's is logical. The opposition is the central debate of the field and goes unmentioned.

Incommensurability is the absence of a common measure between competing theories. Where it holds, a decisive test between them cannot be constructed. It is not raised.

The replication crisis is the finding that a large share of published results fail to reproduce. It is the empirical state of refutation in practice. Accounts of falsification do not mention it.

Pre-registration fixes hypothesis and analysis before data are seen. It exists because unregistered tests can be adjusted until they pass. It is the practical defence of falsifiability.

P-hacking is the selective analysis that avoids an unwelcome refutation. Each individual decision looks defensible. Together they turn a failed test into a successful one.

HARKing is formulating the hypothesis after the results are known and presenting it as a prediction. It converts an exploration into a confirmation. It makes refutation structurally impossible.

Statistical power is the probability of detecting an effect that is really there. Underpowered studies cannot refute anything. A field of underpowered studies produces non-refutations that mean nothing.

Publication bias is the preference for positive findings over negative ones in publication. Refutations are the least publishable result there is. The literature therefore contains fewer refutations than were produced.

Peer review is the institutional check before publication. It is the point at which refutations are filtered as much as errors. Its role in what gets refuted is not examined.

Confirmation bias is the researcher's tendency to favour supporting evidence. It is the psychological counterpart to every structural evasion on this page. Method exists because the disposition does.

Triangulation secures findings through several independent methods. It is a substitute for decisive refutation. Agreement across methods that fail differently is the strongest practical evidence available.

Epistemic injustice is the distorted assessment of testimony because of who is giving it. Refutations from the wrong quarter are weighed differently from refutations from the right one. It decides which challenges are heard.

A causal graph specifies assumed cause and effect relations explicitly, so that what a test can and cannot show becomes visible. It makes the auxiliary assumptions of a study inspectable. It is the modern answer to Duhem-Quine.

A claim that no observation could contradict is unsupported by every observation that appears to support it. Naming what would count against your claim is what makes it a claim rather than a position. It is also rare enough to be distinctive.