How a New Result Is Established
This page assembles what the rest of this group covers into one path. Nothing in it is a formality: each stage exists because a specific class of error survived the stage before it, and the order reflects which errors are cheapest to catch first.
The Sequence
The first filter is internal. In a large collaboration this is a formal analysis review by members who did not do the work but do have the raw data and the code; in a small group it is whatever scrutiny colleagues apply before anything leaves the building. This stage catches arithmetic, coding errors, missing cross-checks and overstated conclusions, and in large experiments it is usually the hardest review the result will face.
The second is publication of a preprint, which exposes the claim to everyone with the relevant expertise at once. In physics this has been the normal first public step since 1991, and community reaction typically arrives within days. The third is peer review at a journal, where two or three specialists assess method and logic without access to the data - a real filter for competence and clarity, and no filter at all for whether the measurement is right.
The fourth is independent replication, and it is the one that changes the status of the claim rather than its presentation. A different group, different apparatus and different analysis team either obtains a compatible result or does not. This is the only stage that can catch a systematic peculiar to one instrument, which is the dominant way confident results turn out wrong.
The fifth is inclusion in the reference compilations - the Particle Data Group tables, CODATA, the review literature - which is where a result stops being a paper and becomes a number other people use without re-deriving it. The sixth and slowest is the textbook, which trails the research front by years and sometimes decades. A result in a textbook has passed every filter; a result at the preprint stage has passed one. Both may be correct, and they do not carry the same weight, and the difference is the number of ways each has been checked.
What a Claim Has to Carry
For a claim to enter this sequence at all, it has to be the kind of thing the sequence can act on. Five elements do that work. A measured quantity, stated in units with a number attached. An uncertainty with its confidence level, because without one there is nothing to compare against. A described apparatus, in enough detail that someone else could build an equivalent. An error budget listing what limits the result. And the conditions under which the measurement holds.
With those five present, a claim is testable by people who have never spoken to its authors, and that is what makes the whole machinery apply to it. Without them, a claim cannot be confirmed even by someone who wants to confirm it, and it therefore does not progress through the filters at all - it stays where it is indefinitely, which is a weaker position than having been tested and found wrong.
This is worth stating in the positive. A specific falsifiable claim with a published method gets resolved fast, and the resolution is sometimes favourable. The LK-99 superconductor claim of 2023 was settled in about four weeks because the recipe was published; had the material been what was claimed, the same four weeks would have confirmed it, from laboratories on three continents. Publishing enough detail to be refuted is also publishing enough detail to be vindicated.
The corresponding failure is not dishonesty but vagueness. A description written to avoid being wrong - no number, no uncertainty, no conditions, a mechanism named rather than specified - cannot be attacked and cannot be supported. It buys immunity from refutation at the price of never being able to accumulate confirmation, and in a field where standing comes from independent replication, that trade is a bad one.
Why Some Claims Cost More
The evidence a claim needs is not fixed. It depends on what else would have to be false if the claim were true, which is why two results with identical statistical strength can be treated very differently without anyone being inconsistent.
A measurement refining a quantity nobody has pinned down needs ordinary evidence: a careful experiment, a stated uncertainty, a confirmation. A measurement that would require revising the first or second law of thermodynamics needs a great deal more, because those laws are not assumptions. They are summaries of an enormous number of measurements across chemistry, engineering and physics, every one of which is also evidence about the new claim. Overturning them means explaining all of that too.
The same asymmetry appears in scientific consensus and it works both ways: the Standard Model prediction that neutrinos are massless was well established and it fell, in about three years, to two independent experiments using different methods. The prior was high and the evidence was higher. Nothing about a strong prior makes a position unfalsifiable; it determines how much measurement is required to move it, which is a question with an answer rather than a barrier.
The distinction that does the real work here is between converting energy that is present in the environment and producing energy from nothing. The first is ordinary physics and the whole of renewable engineering does it - a photovoltaic cell, a thermoelectric element and a piezoelectric harvester all take an existing flux and convert part of it, and the open questions are about efficiency, power density and cost. The second contradicts conservation of energy and is a different kind of claim entirely. Free energy sets out that distinction in detail, and it is the single most useful thing to establish about any ambient-energy proposal before asking anything else about it.
Where a Research Programme Stands
A research programme and an established result are different objects, and the most useful thing anyone presenting a programme can do is say which one they have. A programme has a hypothesis, an apparatus under development, measurements in progress and an expectation. An established result has been measured, published, replicated and compiled. The first can be entirely legitimate and is simply at an earlier stage, and describing it accurately costs nothing it would otherwise gain.
The neutrinovoltaic research pursued by the Neutrino Energy Group in Berlin is a programme in this sense, and the sequence above is the path any programme follows. The questions that advance it are the ordinary ones: what quantity is measured, with what traceable instrument, with what uncertainty, under what conditions, and who independently has obtained a compatible number. These are the same questions that KATRIN answers about the neutrino mass and that IceCube answers about astrophysical neutrino flux, and they are the questions this library applies throughout.
There is also a reason an ambitious programme benefits from this standard rather than suffering under it. A field attracting both serious work and unserious claims cannot be sorted out by assertion from either side, because assertion is exactly what both produce. It can be sorted out by measurement, and the party with a real result is the one that gains when the standard is applied, since the standard is the only mechanism that distinguishes them.
Which leaves a practical test, applicable to any claim in this library's subject matter and symmetrically to all of them. Is there a number with an uncertainty? Is there an apparatus described well enough to rebuild? Is there a statement of what would show the claim to be wrong? And has anyone unconnected to the originators measured the same thing and said what they got? Four questions, no specialist knowledge required, and the answers locate any claim in the sequence without needing anyone's permission.
Frequently asked questions
What is the sequence by which a result becomes established?
Internal review, preprint publication, journal peer review, independent replication, inclusion in reference compilations such as the Particle Data Group tables or CODATA, and finally the textbook. Each stage catches a different class of error, and the order reflects which errors are cheapest to catch first.
Which stage matters most?
Independent replication, because it is the only one that can catch a systematic error peculiar to one apparatus, and that is the dominant way confident results turn out wrong. Peer review filters competence and clarity; it cannot determine whether the measurement is correct.
What does a testable claim need to carry?
A measured quantity in units, an uncertainty with its confidence level, an apparatus described well enough for someone to build an equivalent, an error budget showing what limits the result, and the conditions under which the measurement holds. With those five, strangers can test it.
Why do some claims require more evidence than others?
Because the evidence needed depends on what else would have to be false. A claim requiring revision of thermodynamics has to explain the enormous body of measurement those laws summarise, all of which is also evidence about the new claim. A strong prior sets how much measurement is required, not an impassable barrier.
What is the difference between converting ambient energy and free energy?
Converting energy already present in the environment is ordinary physics, and photovoltaic, thermoelectric and piezoelectric devices all do it; the open questions concern efficiency, power density and cost. Producing energy from nothing contradicts conservation of energy and is a different kind of claim entirely.
