Week 9: The null result

Why "we found nothing" is the hardest sentence in science to publish, and what a discipline loses when it only keeps its positive results

19 April 2027

The file-drawer problem: a null result — the hypothesis wasn’t supported — is exactly as informative as a positive one, and gets published at a fraction of the rate, because journals and researchers both prefer a story with a finding in it. The consequence isn’t just missing papers; it’s a published literature that’s a biased sample of reality, weighted toward whatever happened to work in whichever lab tried it first.

This is the first week where “leaving something out” isn’t a craft choice anyone’s proud of — it’s closer to a structural failure the field is still arguing about how to fix (preregistration, registered reports, mandatory null-result archives). The lecture treats it as the semester’s harshest test of the course’s own thesis: sometimes what’s missing isn’t a considered cut at all, just an absence with a cost nobody’s paying attention to.

Outline#

• the file-drawer problem and why it’s worse than it sounds
• preregistration and registered reports as attempts to force the null result back into the record
• when “what’s missing” stops being a design choice and starts being a bias