“A scientific study proves something is true.”

• In everyday usage, 'proves' often means 'provides strong evidence for,' and well-designed studies (large, pre-registered, randomized controlled trials) can establish causal conclusions with very high confidence — e.g., the evidentiary base behind smoking–lung cancer causation or vaccine efficacy trials.
• Some scientific claims are effectively settled by observation: a single verified observation can definitively establish an existence claim (a new element, particle, or species) or definitively falsify a universal claim by counterexample, which is closer to genuine 'proof.'
• Regulatory and legal systems treat certain study results as dispositive: agencies such as the FDA and EMA approve drugs on the basis of 'substantial evidence' from adequate and well-controlled trials, operationally treating such findings as established fact.
• Formal deductive proof does exist inside science's mathematical and logical components (theorem proving, derivations), so the word 'proof' is not wholly alien to scientific practice.

• Empirical science is inductive and probabilistic: statistical inference quantifies uncertainty and can reject or fail to reject hypotheses, but cannot deductively establish truth. The American Statistical Association's 2016 statement explicitly warns that p-values do not measure the probability that a hypothesis is true and should not be treated as proof.
• Philosophy of science since Popper holds that theories can be corroborated or falsified but never conclusively verified; all empirical conclusions remain provisional and open to revision by new data.
• A *single* study is especially weak grounds for truth claims. The Open Science Collaboration (2015) replicated only about 36% of 100 published psychology findings, and Camerer et al. (2018) replicated roughly 13 of 21 high-profile social science experiments in Nature/Science, with effect sizes about half the originals.
• Ioannidis (2005) argued formally that, given small sample sizes, small effects, flexible analysis, and publication bias, most published research findings are likely false — directly undercutting 'a study proves it.'
• The U.S. National Academies' 2019 consensus report 'Reproducibility and Replicability in Science' emphasizes that confidence in a result comes from accumulated, replicated, and converging evidence, not a single publication.
• Numerous individual studies that were widely reported as 'proof' were later retracted or overturned (e.g., the 1998 Wakefield MMR–autism paper, retracted by The Lancet in 2010), demonstrating that publication is not proof.
• Observational studies are vulnerable to confounding, selection bias, and reverse causation, so they establish association rather than proven causation; even the Bradford Hill criteria are framed as guides to causal inference, not proof.
• The Royal Society's founding motto 'Nullius in verba' ('take nobody's word for it') reflects the institutional norm that claims are held open to independent testing rather than accepted as proven on authority.

90/100 — the lab’s own confidence in the score above, not a second reading of the claim.

Open Science Collaboration, Science (AAAS)journal · 95 · Camerer et al., Nature Human Behaviourjournal · 93 · Ioannidis, PLOS Medicinejournal · 90 · Wasserstein & Lazar, The American Statistician (American Statistical Association)journal · 95 · National Academies of Sciences, Engineering, and Medicineedu · 96 · Stanford Encyclopedia of Philosophyedu · 92 · Understanding Science, University of California, Berkeleyedu · 85 · The Royal Societyother · 88 · U.S. Food and Drug Administrationgov · 90 · The Lancet (retraction of Wakefield et al. 1998)journal · 94

• A well-designed scientific study can strongly support a claim by collecting empirical data under systematic methods; regulatory agencies such as the FDA rely on adequate and well-controlled studies as part of determining whether drugs are effective and safe enough for approval.
• When findings are replicated by independent researchers, combined in systematic reviews or meta-analyses, and supported by plausible mechanisms, scientific studies can justify very high confidence in a claim.

• The wording 'proves something is true' is too absolute: empirical science generally produces provisional conclusions with quantified uncertainty, not mathematical proof.
• A single study can be wrong because of bias, confounding, small sample size, selective reporting, flawed measurement, p-hacking, or random chance; Ioannidis' widely cited PLOS Medicine paper explains why many published research findings can be false.
• Replication failures show that some published findings do not hold up when retested; the Open Science Collaboration reported that many psychology studies produced weaker or non-significant effects in replication attempts.
• The National Academies distinguishes reproducibility and replicability and emphasizes that non-replication can arise from multiple causes, including methodological problems and ordinary scientific uncertainty.
• Therefore, a more accurate statement is: a scientific study can provide evidence for a claim, sometimes very strong evidence, but it does not by itself prove the claim is definitively true.

91/100 — the lab’s own confidence in the score above, not a second reading of the claim.

U.S. Food and Drug Administrationgov · 95 · PLOS Medicinejournal · 88 · Sciencejournal · 90 · National Academies of Sciences, Engineering, and Medicineother · 94 · Naturejournal · 86

• In colloquial language and popular media, the word 'proof' is frequently used as a simplified shorthand to mean 'provides strong empirical evidence for.'
• When highly rigorous, mathematically grounded studies in fields like physics or chemistry repeatedly yield the same results with extremely high statistical confidence, the findings are often treated as factual reality or 'truth' for all practical purposes.

• In the philosophy of science, the scientific method relies on inductive reasoning and falsifiability. Science gathers empirical evidence to support or refute hypotheses, but it does not 'prove' anything with absolute mathematical or logical certainty.
• A single scientific study is rarely conclusive. The scientific community relies on consensus built over time through meta-analyses, systemic reviews, and repeated experiments, rather than taking a single study as absolute truth.
• The ongoing 'replication crisis' in science, particularly in medicine and psychology, has demonstrated that a significant percentage of single-study findings cannot be reproduced when experiments are repeated by independent researchers.
• Individual studies are frequently subject to limitations such as small sample sizes, publication bias, methodological flaws, confounding variables, and statistical anomalies like p-hacking, which can lead to false conclusions.

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

PLOS Medicinejournal · 95 · Naturejournal · 98 · Forbesnews · 85

• In everyday language people often use “prove” to mean “provide strong evidence for,” and **well-designed experimental studies** can provide compelling, quantitative support for specific hypotheses in a defined context, especially when they show large effects with low risk of bias and clear mechanisms.[1][8]
• Philosophers of science note that in some decision-making contexts it can be *rational* to treat a single, well-conducted study as **decisive evidence** for or against a hypothesis—for example, when further data are very costly and the study’s signal-to-noise ratio is high.[3]
• Under Popperian falsification, a single counterexample can logically **falsify a universal claim** (e.g., one black swan refutes “all swans are white”), meaning that a single study can definitively show that a universal statement is *not* true if the methods and observations are sound.[6][8][12]
• In practice, regulatory and clinical decisions are sometimes based heavily on one pivotal trial (e.g., a large randomized controlled trial), which functionally treats that single study as “proof enough” for action, even though the underlying philosophy acknowledges uncertainty.[1][8]

• Authoritative discussions of how science works emphasize that **science can neither prove nor disprove hypotheses in an absolute, final sense**; instead, scientists evaluate hypotheses based on accumulating supporting and refuting evidence.[1][12]
• Philosophy of science sources explicitly state that observational evidence **cannot prove or verify** a universal hypothesis; it can reliably *refute* hypotheses but not establish them as certainly true, because further observations might always contradict the claim.[8][12]
• Modern treatments of scientific evidence stress that **no single study is enough to settle a question**, and that science depends on replication by independent teams, using different methods and populations, before strong confidence or consensus is warranted.[4][5][9][15]
• The widely discussed **replication/reproducibility crisis** shows that a substantial share of published findings fail to replicate when other researchers repeat the work, demonstrating that many conclusions based on single studies were not robust and could not be treated as proved.[2][4][5][11][15]
• Analyses of replication efforts recommend focusing on **distributions of effect sizes** and multiple studies rather than trying to obtain closure about whether an effect is true or false from a single experiment, warning against the “solo experiment” mentality.[4][7][15]
• Philosophical and methodological work highlights that single case studies or experiments are often a **poor basis for establishing general claims across populations**, and while a single case can refute a hypothesis, it is “not even in principle sufficient” to confirm broad quantitative theses.[13][14]

93/100 — the lab’s own confidence in the score above, not a second reading of the claim.

National Academies of Sciences, Engineering, and Medicinegov · 95 · FORRT (Framework for Open and Reproducible Research Training)edu · 85 · PhilArchiveedu · 80 · Psychological Science (via Sci-Hub)journal · 90 · CASRAIedu · 80 · Internet Encyclopedia of Philosophyedu · 85 · Replication Index (U. of Toronto-affiliated blog)edu · 75 · National Library of Medicine / NIHgov · 95 · Science Insightsother · 70 · Harvard Universityedu · 90 · Medical Xpressnews · 80 · University of Michiganedu · 85

• Rigorous, replicated studies in mature fields (physics, chemistry, parts of medicine) produce findings treated as established facts for practical purposes, such as germ theory, orbital mechanics, or vaccine efficacy.
• Peer review, statistical testing, and independent confirmation can make a body of studies highly reliable evidence that something is true beyond reasonable doubt in ordinary language.

• Science does not deliver absolute proof; hypotheses are supported or weakened by evidence and remain open to revision (falsifiability). A single study almost never suffices.
• Publication bias, p-hacking, small samples, and the replication crisis mean many published results do not hold; one paper cannot be treated as proving a claim true.
• The wording equates empirical evidence with logical/mathematical proof, which overstates certainty and is a common source of public misunderstanding.

78/100 — the lab’s own confidence in the score above, not a second reading of the claim.

PLOS Medicinejournal · 90 · Science (AAAS)journal · 95