Preregistering a study should be normal, not heroic

Preregistration means writing down, before collecting data, what you will test and how you will analyse it, and publishing that plan. It sounds modest, but it addresses a real weakness: when researchers decide their analysis after seeing the data, it becomes easy, usually without bad intent, to find a pattern that looks significant and is really chance.

We think preregistration should be the default for studies that claim to confirm a hypothesis. It does not stop exploration; exploratory analyses are valuable and should simply be labelled as exploratory. It makes the difference between "we predicted this" and "we noticed this" visible to readers.

The honest objection is rigidity: real research changes direction, and a plan can feel like a straitjacket. Our answer is that deviations are fine if they are reported and explained.

What would change our mind: strong evidence that preregistration produces no measurable improvement in the reliability of findings, or that it deters more good research than it protects.

The rule of thumb for readers: when you see a striking result, ask whether the analysis plan was fixed beforehand.