Peer-reviewed journals — as well as researchers and their funders — must take responsibility for improving the reproducibility of published results (see Nature 533, 452–454; 2016).

I suggest that journals should be required to sign a global statement indicating that, to the best of their knowledge, the data that they publish are reproducible. This statement would be collaboratively formulated by the editors-in-chief in accordance with recommendations from the International Committee of Medical Journal Editors and guidelines proposed by the US National Institutes of Health, Nature and Science (see Nature 515, 7; 2014 and go.nature.com/29bxphv).

Journals would then publish only papers that are accompanied online by full experimental protocols, raw data and source code, as in the Protocol Exchange repository (www.nature.com/protocolexchange). For manuscripts containing statistical analyses, journals should peer review only those papers that use statistics environments based on source code, enforcing the ban on 'point-and-click' statistical software (see go.nature.com/29pdpcl).