Much of what scientists learn in the laboratory never reaches the literature. Experiments that fail to confirm a hypothesis are often left in drawers, distorting the evidence base and wasting resources as others repeat them.

Journals, funders and institutions all have a part to play. Registered reports, in which methods are peer reviewed before results are known, remove the incentive to bury null findings. Dedicated article types and data repositories make it easy to share them.

Culture matters most. Supervisors who treat a well-designed experiment with a negative outcome as a success will raise a generation of more rigorous scientists.