Evaluating Science's open-data policy

See allHide authors and affiliations

Science  18 Aug 2017:
Vol. 357, Issue 6352, pp. 654
DOI: 10.1126/science.aan8158

Science must re-evaluate its open-data policy after retracting a controversial study on microplastics and fish in May (“Editorial Retraction,” J. Berg, 26 May, p. 812 and “Fishy business,” M. Enserink, News Feature, 24 March, p. 1254). The only computer containing the study's raw data was allegedly stolen and no backups existed on another machine or an online repository. Many are left wondering how this could happen in an era of cloud computing and open data.

Science currently has a two-step editorial policy to promote data availability and research transparency (1). First, “before publication, large data sets must be deposited in an approved database and an accession number…included in the published paper.” Second, “after publication, all data [and code] necessary to understand, assess, and extend the conclusions of the manuscript must be available to any reader of Science.” Unfortunately, two loopholes weaken the effectiveness of this policy. First, it is unclear why “small data sets” are exempt from archival before publication given that they are simple to share and equally important in verifying research results (as the above case demonstrates). Second, research shows that individual scientists cannot be trusted to reliably preserve their data and share them upon request (2, 3). Compounding this problem is the fact that publishers seldom state the consequences of breaching their policies. For example, does failure to provide raw data post-publication lead to the automatic retraction of the paper? The solution to avoid data disappearing is simple: Journals must mandate and enforce data archival on a recognized, online repository at the time of submission. Only editors and reviewers would have access during peer review; the data would be made generally accessible upon publication.

Publishing verifiable research is a tenet of scientific progress and, ultimately, journals are responsible for guaranteeing compliance with their open-data policy. At a minimum, this responsibility involves a cursory check of the underlying data and ensuring that all data are available for reviewers to assess (4). Science publishes many papers describing major breakthroughs, but these extraordinary claims must be supported by extraordinary evidence. This includes, first and foremost, a complete and understandable data set that is open to reviewers and, ultimately, becomes open to scientists and the public.


Navigate This Article