A finding that cannot be by an independent team is not yet knowledge.
Large projects in psychology, medicine and economics have failed to reproduce between a third and two thirds of published results.
The reaction to those figures has been revealing in itself.
Some researchers treated them as evidence of fraud, which they largely are not.
Deliberate exists and is rare; the ordinary causes are structural and far harder to remedy.
A career is built on publication, and journals have historically published positive, novel and surprising findings.
Every therefore points towards analysing data until something reaches .
This need not involve dishonesty, since a researcher may sincerely believe each analytical choice was the right one.
The remedy that has worked best is also the least glamorous.
requires the hypothesis and the analysis to be recorded before the data are collected.
Registered reports go further, with journals accepting the paper on the strength of the design rather than the outcome.
In fields that adopted them, the proportion of published rose sharply, which is exactly what should happen.
Replication itself remains badly funded, because repeating somebody else's study advances nobody's career.
Several funders now reserve a fixed share of grants for , which is a cheap correction.
A subtler danger has emerged as the reform matures.
Fields that report low rates appear untrustworthy, while fields that have never checked appear reliable.
Transparency thus carries a reputational cost that avoids.
Any honest assessment of a discipline should begin by asking whether it has looked.