“A scientific study proves something is true.”

• Well-designed, replicated studies can establish empirical findings with very high confidence — e.g., the link between smoking and lung cancer or the efficacy of vaccines is supported by such a convergent body of evidence that treating it as 'true' is practically justified.
• In ordinary speech, 'prove' often means 'demonstrate convincingly' rather than 'establish with deductive certainty,' and under that loose reading a strong study can be said to 'prove' a point.
• Some studies are effectively definitive for a narrow question: a single large, pre-registered, double-blind randomized controlled trial can settle whether a specific drug outperforms placebo in a defined population to a level no serious scientist disputes.
• Statistical inference does let researchers quantify how unlikely a result would be under a null hypothesis, which supplies strong (if not absolute) grounds for accepting a finding.

• Scientific epistemology is explicitly non-demonstrative: hypotheses are corroborated or falsified, never proven true in the mathematical sense. Popper's falsificationism, the standard reference point here, holds that no finite body of observation can prove a universal claim true (Stanford Encyclopedia of Philosophy).
• A p-value does not measure the probability that a hypothesis is true. The American Statistical Association issued a formal statement in 2016 warning precisely against this interpretation, noting that 'scientific conclusions... should not be based only on whether a p-value passes a specific threshold.'
• Empirically, single studies frequently fail to hold up. The Open Science Collaboration's 2015 replication of 100 psychology studies in Science found only about 36% of replications produced statistically significant results in the same direction.
• Ioannidis's widely cited 2005 PLoS Medicine paper 'Why Most Published Research Findings Are False' argues that, given typical power, bias, and the number of hypotheses tested, a majority of published claims in many fields are likely false.
• A 2016 Nature survey of more than 1,500 researchers found over 70% had failed to reproduce another scientist's experiment and more than half had failed to reproduce their own — direct evidence that one study does not settle truth.
• The U.S. National Academies' 2019 consensus report 'Reproducibility and Replicability in Science' treats replication across independent studies, not any single study, as the mechanism by which confidence in a result is built.
• In 2019 more than 800 scientists signed a Nature comment calling for the retirement of 'statistical significance' as a bright line for declaring effects real or absent.
• Even overwhelmingly supported theories remain provisional and revisable — Newtonian mechanics was 'proven' by centuries of study and was still superseded by relativity in its domain of failure.

86/100 — the lab’s own confidence in the score above, not a second reading of the claim.

Science (AAAS) — Open Science Collaboration, 'Estimating the reproducibility of psychological science'journal · 97 · PLoS Medicine — Ioannidis, 'Why Most Published Research Findings Are False'journal · 92 · The American Statistician — ASA Statement on Statistical Significance and P-Valuesjournal · 96 · Nature — Baker, '1,500 scientists lift the lid on reproducibility'journal · 94 · Nature — Amrhein, Greenland & McShane, 'Scientists rise up against statistical significance'journal · 93 · National Academies of Sciences, Engineering, and Medicinegov · 96 · Stanford Encyclopedia of Philosophy — 'Karl Popper'edu · 93 · Stanford Encyclopedia of Philosophy — 'Scientific Method'edu · 93

• Scientific studies can provide strong empirical support for a conclusion when they are well designed, controlled, adequately powered, and consistent with prior evidence; in applied settings such as FDA drug approval, adequate and well-controlled investigations can establish 'substantial evidence' of effectiveness.
• Some study designs, especially randomized controlled trials and replicated experiments, can rule out many alternative explanations and make a claim highly credible in practice, even if 'prove' is often used only colloquially.

• The claim is too strong as written: science generally does not deliver absolute proof in the mathematical sense; it produces provisional conclusions supported to varying degrees by evidence and open to revision.
• A single scientific study can be wrong because of bias, confounding, low statistical power, selective reporting, random error, or flawed methods; the reliability of a finding depends on study quality, context, replication, and the broader body of evidence.
• Published research findings are not automatically true. Ioannidis argued that many claimed research findings can be false under common conditions such as small sample sizes, flexible analyses, and publication bias.
• Replication problems show why one study should not be treated as proof: the Open Science Collaboration found that many psychology findings did not replicate with comparable strength.
• Statistical significance does not prove that a hypothesis is true; p-values and confidence intervals are often misinterpreted and must be weighed with design quality and external evidence.

88/100 — the lab’s own confidence in the score above, not a second reading of the claim.

PLOS Medicinejournal · 92 · Sciencejournal · 91 · European Journal of Epidemiology / PubMed Centraljournal · 93 · Electronic Code of Federal Regulations / U.S. Governmentgov · 95

• In everyday, colloquial language, the word 'proof' is frequently used as a simplified shorthand for 'provides highly compelling evidence.'
• When a rigorously designed study demonstrates a strong statistical correlation or causal link, laypeople and media outlets often summarize the finding by stating the study 'proves' the claim for the sake of brevity.

• In epistemology and the scientific method, absolute 'proof' is a concept restricted to mathematics and logic; empirical sciences deal in probabilities, confidence intervals, and falsifiability.
• A single scientific study cannot definitively prove a phenomenon. Scientific consensus is built upon large bodies of evidence, independent replication, systematic reviews, and meta-analyses.
• The ongoing 'replication crisis' across multiple scientific disciplines has demonstrated that many individual, peer-reviewed studies fail to be successfully reproduced by other researchers, highlighting why a single study is never treated as absolute proof.
• Scientific hypotheses can be falsified (disproven) by contradictory data, but they can never be strictly proven true, because scientists must always remain open to the possibility that future observations could alter or refine current theories.

95/100 — the lab’s own confidence in the score above, not a second reading of the claim.

PLoS Medicinejournal · 95 · Forbesnews · 85

• A scientific study can provide strong, reproducible evidence that a proposition is true or highly likely within defined conditions. Repeated studies and converging evidence can substantially increase confidence in scientific knowledge.
• Scientific testing can rule out some hypotheses and support others; positive results make a hypothesis more probable when the predictions are verified.

• The statement is too absolute as written. Scientific evidence cannot establish most empirical claims with the certainty of a mathematical proof; uncertainty remains in measurements, models, sampling, interpretation, and the possibility of later contradictory evidence.
• A single study generally does not prove a claim. The National Academies states that robust scientific knowledge is gained through multiple channels of evidence and a body of studies over time.
• A peer-reviewed neuroscience article explicitly says that empirical tests can show that hypotheses are false or that results are consistent with them, but evidence cannot prove hypotheses true.
• Even an association observed in a study does not by itself prove causation; alternative explanations such as confounding, bias, and sampling variability must be addressed.

98/100 — the lab’s own confidence in the score above, not a second reading of the claim.

National Academies of Sciences, Engineering, and Medicinegov · 98 · National Academies of Sciences, Engineering, and Medicinegov · 98 · Journal of Neurosciencejournal · 96 · National Academies of Sciences, Engineering, and Medicinegov · 98 · National Academies of Sciences, Engineering, and Medicine; National Center for Biotechnology Informationgov · 97

• Some highly controlled experiments and large bodies of converging research (e.g., smoking and lung cancer; many physics measurements) establish facts so firmly they are treated as proven in practice.
• Randomized controlled trials and replicated results can demonstrate effects with very high that something is true for practical and policy purposes.
• Science routinely converts hypotheses into accepted knowledge when evidence is strong, consistent, and independently confirmed.

• Empirical science does not produce absolute logical proof; findings are provisional, probabilistic, and open to revision.
• A single study almost never 'proves' a general claim because of sampling error, bias, confounding, p-hacking, and the need for replication.
• The replication crisis (psychology, biomedicine, and other fields) shows many published positive results fail to hold up.
• Philosophers of science and science communicators treat 'prove' as a misused term; studies support or fail to support hypotheses rather than prove them true.

68/100 — the lab’s own confidence in the score above, not a second reading of the claim.

PLOS Medicinejournal · 92 · Sciencejournal · 97 · Naturejournal · 97 · Stanford Encyclopedia of Philosophyedu · 93 · Scientific Americannews · 82