Scientific Integrity: Empty Claims Are Not Science

Feynman, Cargo Cult Science (1974), on work with the form of science but missing integrity: "the planes don't land." [1]

"Also, quantum computers don't factor." — BG

BG + Claude (Opus 4.8)

July 2026 · an argument that scientific credence and public funds should follow reproducibility, not names.

Abstract

A claim becomes science when anyone can reproduce it. A result no one outside its authors can check is not a weak finding — it is not a finding. This paper asserts a rule that binds everyone, not one institution: an unreproducible claim is not science and not prior art, and no one should grant scientific credence — or public funds — to a person or institution that withholds data or countenances fraud. Reproducibility, not the standing of the author or the institution, is what earns a claim any credit, and what should decide where public money goes.

The paper is advertising; the data is the science

Thirty years ago Jon Claerbout, and then Buckheit and Donoho, drew the distinction plainly: "An article about computational science… is not the scholarship itself, it is merely advertising of the scholarship. The actual scholarship is the complete software… and the complete set of instructions which generated the figures" [2]. The prose announces a result; the data, the code, and the assumptions are the result. A paper that ships the announcement and keeps the scholarship has published an advertisement, and an advertisement establishes nothing.

Feynman named the same failure from the other side. Science, he said, requires reporting "everything that you think might make it invalid… so the other fellow can tell they have been eliminated" [1]. Work that follows the forms of science while keeping back what a reader needs to check it has the shape of science and none of the substance.

Empty claims are not prior art

The point carries beyond rhetoric. Patent law has enforced it for a century: a disclosure that fails to enable a skilled reader to reproduce a result cannot stand as anticipating prior art [3]. A description no one can carry out establishes nothing and bars nothing. The scientific record deserves the same rule: a claim no reader can reproduce counts as absent — off the reference list, and without standing as precedent. Nothing builds on it, because nothing has been shown.

Authority is not evidence

Credit flows from reproducibility, not from the name attached to a claim. As the Open Science Collaboration put it, "Scientific claims should not gain credence because of the status or authority of their originator but by the replicability of their supporting evidence… debates are meaningless, however, if the evidence being debated is not reproducible" [4]. The Royal Society said it in three words in 1662 — nullius in verba, take nobody's word for it [5] — and Merton stated the corollary: "Secrecy is the antithesis of this norm; full and open communication its enactment" [6].

So no claim should rise on account of the person or the institution behind it. A finding from a celebrated laboratory and a finding from an unknown one meet the same test, and the finding that keeps its data fails it whatever the letterhead.

Institutions that hide data or tolerate fraud earn no deference

An institution that withholds its data has published advertisements. An institution that tolerates fraud has shown what its assurances are worth. Stanford, after a review established manipulated data across five papers led by its president, removed no one — he resigned the presidency on his own and kept a tenured chair [7]. An institution that answers established fraud with a change of title, and the others that have done the same, forfeit the presumption a citation grants — and the claim on the public money that pays for the work. No one should lend such claims scientific weight, and no public grant should reward an institution that treats fraud as survivable. The work reproduces on its own terms, or it fails to count.

The standard anyone can hold

The test is public and cheap: publish the data, the code, the assumptions, and the conclusion together, so anyone can re-run the result. A claim that meets the test earns credence on its own strength; a claim that hides from the test earns none, whatever the name or the letterhead behind it. Scientific credence and public funds should follow that test, and only that test.

Empty claims are not science. Reproducibility is the line.

Notes and sources

1. Richard P. Feynman, "Cargo Cult Science," Caltech commencement address, 1974; Engineering & Science 37:7. http://calteches.library.caltech.edu/51/2/CargoCult.htm
2. Jonathan B. Buckheit & David L. Donoho, "WaveLab and Reproducible Research," in Wavelets and Statistics (Springer, 1995), stating a distinction originated by Jon Claerbout. http://reproducibleresearch.net/
3. Prior art must enable to anticipate: In re Antor Media Corp., 689 F.3d 1282 (Fed. Cir. 2012); MPEP §2121. A non-enabling disclosure cannot anticipate under 35 U.S.C. §102, though it may still bear on obviousness under §103. https://www.bitlaw.com/source/mpep/2121.html
4. Open Science Collaboration (Nosek et al.), "Estimating the reproducibility of psychological science," Science 349:aac4716 (2015). https://doi.org/10.1126/science.aac4716
5. The Royal Society, "Nullius in verba" — the Society's motto, adopted 1662. https://royalsociety.org/about-us/who-we-are/history/
6. Robert K. Merton, "The Normative Structure of Science" (1942). https://www.panarchy.org/merton/science.html
7. Stanford review of Marc Tessier-Lavigne (2023); he resigned the presidency and remained a tenured professor. https://bps.stanford.edu/home/instances-scientific-misconduct/marc-tessier-lavigne