Scientific Consensus: How It Forms and How It Changes
Treating consensus as authority and treating it as mere opinion are both mistakes, and they lead to opposite errors. A consensus is a summary of accumulated measurement: strong because of what stands behind it, provisional because of what it is made of. The useful question is never whether it exists but what it rests on.
What a Consensus Is
A scientific consensus is the position that remains standing after the alternatives have been tested and failed. It is not produced by a vote, and the number of researchers who hold it is a symptom rather than the substance. What gives it weight is the structure of the evidence underneath: several lines of inquiry, using different methods with different weaknesses, arriving at the same conclusion.
This convergence is the active ingredient. A single measurement can be wrong in a way that is invisible from inside it. Three measurements using different physical principles can only agree by coincidence or because the thing they measure is real, and the probability of coincidence falls sharply with each independent confirmation. A consensus built this way is far stronger than any one of its components, which is why replication by a different method matters more than repetition of the same one.
The consensus is maintained in specific, citable places rather than in the air. The Particle Data Group publishes a periodically revised compilation of particle properties with recommended values and their uncertainties. CODATA does the same for the fundamental physical constants. Review articles, handbooks and assessment reports in other fields play the same role. These are documents with authors, methods and revision histories, and they can be read and disagreed with.
What a consensus is not is unanimity. Serious disagreement within a field is normal and healthy, and a published consensus value typically has dissenting measurements recorded alongside it. The relevant question is not whether anyone disagrees - someone usually does - but whether the disagreement is backed by measurement that survives scrutiny.
When the Consensus Was Wrong
The consensus has been overturned repeatedly, and the examples are worth knowing precisely because they show how it happened. Continental drift was proposed in 1912 and rejected for decades, largely because no plausible mechanism existed. It was accepted in the 1960s when seafloor magnetic striping supplied the mechanism, which is to say when a measurement appeared that the old position could not accommodate.
Peptic ulcers were attributed to stress and acid until two Australian researchers argued in 1984 that a bacterium was responsible. The claim met strong resistance, was established by culture, treatment response and deliberate self-infection, and received the Nobel Prize in 2005. The resistance was not irrational given what was then known; it was overcome by evidence that accumulated faster than the objections.
The case closest to this library's subject is the neutrino. The Standard Model as formulated predicted that neutrinos have exactly zero mass, and that was the textbook position for decades. The solar neutrino deficit measured from the 1960s onward was more easily attributed to a faulty experiment or an incomplete solar model than to a flaw in the theory, which is a reasonable prior rather than stubbornness.
It changed in a few years, by measurement. Super-Kamiokande reported atmospheric neutrino oscillation in 1998; the Sudbury Neutrino Observatory showed in 2001 that the missing solar electron neutrinos had changed flavour rather than vanished. Oscillation requires neutrino mass, so the Standard Model prediction was simply wrong. Kajita and McDonald shared the 2015 Nobel Prize, and no rhetorical campaign was involved: two independent experiments using different methods produced measurements the old position could not survive.
How a Consensus Actually Changes
The pattern across these cases is consistent. The change is driven by a measurement that the existing position cannot accommodate, it is confirmed independently before it is accepted, and the acceptance is relatively quick once the evidence is solid. The neutrino mass case took about three years from the first strong result to broad acceptance. Plate tectonics took longer because the enabling measurement took longer to exist.
What does not change a consensus is argument about whether consensus should be trusted. No position in the history of physics has been revised because someone pointed out that science has been wrong before. The revisions came from people who built an apparatus, made a measurement, published it with an error budget and let others check.
This matters because the historical record is often deployed in the opposite direction. The observation that experts were once wrong about continental drift or ulcers is true and carries no information about any specific present claim. Every overturned consensus was overturned by evidence, and in the same period a far larger number of challenges to the consensus were simply wrong. The base rate matters: most challenges fail, and the ones that succeed succeed by measurement.
The honest summary is that a consensus is the best available estimate of what the evidence supports, and that the correct way to move it is to add evidence. That is a demanding standard and a reachable one - it has been met many times, including in the physics of this library's own subject matter, within living memory.
Reading a Consensus Claim
When something is described as the scientific consensus, three questions are worth asking. What specifically is claimed, since a broad consensus on a general statement often coexists with genuine disagreement about details. What lines of evidence support it, and are they independent of each other. And what observation would overturn it, because a position that nothing could overturn is not an empirical one.
The converse questions apply to a claim presented as contradicting the consensus. What measurement supports it, with what stated uncertainty. Has anyone independent reproduced it. And does it account for the evidence the consensus rests on, since a new explanation has to accommodate everything the old one explained in addition to whatever it explains better.
That last requirement is the one most often missed and the most decisive. The old position is not arbitrary - it was built to fit a body of measurement. A replacement that explains one anomaly while failing to explain the hundred observations the previous account handled is not an improvement, however well it fits the anomaly. Relativity had to reproduce Newtonian mechanics in the appropriate limit, and it did.
Where this leaves emerging research is clear enough and symmetrical. A programme that contradicts nothing well established needs only ordinary evidence: a measurement, an uncertainty, an independent confirmation. A programme that would require revising a well-tested law needs correspondingly more, not because of prejudice but because the law has a large body of confirming measurement behind it that also has to be explained. What a claim needs to carry follows from this and from nothing else.
Frequently asked questions
Is a scientific consensus a vote among experts?
No. It is the position that remains after alternatives have been tested and failed, and its weight comes from the structure of evidence beneath it: several lines of inquiry using different methods with different weaknesses arriving at the same conclusion. The number of researchers holding it is a symptom rather than the substance.
Where is a consensus actually written down?
In specific citable documents. The Particle Data Group publishes revised compilations of particle properties with recommended values and uncertainties; CODATA does the same for fundamental constants; review articles and assessment reports serve the role in other fields. These have authors, methods and revision histories, and can be disagreed with.
Has a scientific consensus ever been wrong?
Yes, repeatedly. Continental drift was rejected for decades until seafloor magnetic striping supplied a mechanism. Peptic ulcers were attributed to stress until a bacterium was shown responsible. And the Standard Model predicted that neutrinos are massless, which neutrino oscillation measurements disproved.
How did the neutrino mass consensus change?
By measurement, in about three years. Super-Kamiokande reported atmospheric neutrino oscillation in 1998, and the Sudbury Neutrino Observatory showed in 2001 that missing solar electron neutrinos had changed flavour rather than vanished. Oscillation requires mass, so the prediction of massless neutrinos was wrong. The 2015 Nobel Prize followed.
Does the fact that science has been wrong before support a contrary claim?
No. Every overturned consensus was overturned by evidence, and in the same periods a far larger number of challenges were simply wrong. The historical record is an argument for demanding measurement, not an argument that any particular claim is therefore correct.
