Null Results, and Why Science Needs Them

The most consequential experiment of the nineteenth century found nothing. Michelson and Morley set out to measure the ether wind and detected no such thing — a null result so stubborn and so precise that it eventually demanded relativity. The episode is taught as a triumph. Its modern descendants, by and large, are not even published.

Publication bias is the polite term: journals prefer positive findings, careers require journals, and so the literature fills with discoveries while the disconfirmations accumulate in file drawers. The arithmetic is corrosive. If twenty labs test a false hypothesis and one gets a fluke positive, the fluke gets published and the nineteen correctives never surface. Entire subfields have been built on exactly this foundation, as the replication crises of psychology, oncology, and economics spent a decade demonstrating.

The waste runs deeper than error. A null result is information — often expensive information — about where the effect is not, and its suppression forces the same dead ends to be rediscovered at every institution too proud to publish the map. Physics learned this lesson institutionally: the null results of dark-matter searches and collider exclusions are published as a matter of course, each one narrowing the space where a theory may still live. The exclusion plot is the field's monument to productive failure.

The remedies are known and slowly spreading — registered reports that accept papers on design before results exist, funders that require outcome reporting, journals of record for negative findings. What lags is the culture: the tenure file still counts discoveries, and no one is feted for the effect they correctly failed to find.

That is a failure of storytelling as much as of incentives. The experiment that finds nothing, done well, is science at its most honest — the discipline paying, in full and in public, the cost of being checkable. A culture that celebrated its null results would not be settling for less. It would be, at last, taking its own epistemology seriously.