Data sharing and data publication

Despite regulation supporting sharing of scientific data and extensive investment in neuroinformatics database infrastructure very little data is actually being shared in most subfields of neuroscience. I will first discuss reasons why data sharing often fails in practice, referring to personal experience as computational neuroscientist and co-editor in chief of Neuroinformatics. I will then argue that the main problem is a lack of clear incentives to share data. An obvious incentive can be authorship, but it is often not practical or appropriate for the scientists who collected the data to become co-authors of original articles that report results derived from the data. Recently, an alternative has become available under the form of data publication. Initially this was mainly seen as a procedure to make publicly released data collections citable by providing a doi. But several journals, including Neuroinformatics, have now introduced the concept of a data article. This is a paper that just describes how the data was collected and how it is made available, but does not include any interpretation or conclusions. The advantages and attractiveness of the data article format will be discussed.