Richard Smith: A better way to publish science

richard_smith_2014Journals have been the main way to publish scientific research for 400 years, and remarkably they still are despite 20 years of the World Wide Web. But it’s becoming increasingly clear both that the journal model is beginning to creak and that better models are appearing. I’m working with F1000, and we propose moving from a world where publishers and editors are involved in making judgments on science to one where publishers simply provide a service to scientists and the funders of science to help them rapidly disseminate their findings. This move should address at least some of the many problems of research and journals.

These are the current problems with journals:

1. Scientific articles rarely include the data that lie behind the paper, a defect that was inevitable in the age of paper but can now easily be remedied.

2. Much research is still never published; only about half of clinical trials are published. Worse, there is bias in what is published with the “positive and exciting” taking precedence over the “negative and dull but important.”

3. Much of the research that is published is of low quality. An important reason for this is that because of the way scientists are assessed, mainly through what and where they publish, it has become more important to publish than do good science.

4. Much research is not reproducible, which is part of the problem of scientists rushing to publish to claim priority and to satisfy systems of assessment.

5. Much research is “wasted,” the Lancet estimates 85% in biomedical science. This is for many reasons, including that questions being answered are already answered, the questions are not of interest, methods are inadequate, bias is not controlled, and research is not published or is not accessible.

6. There is also much waste in the publishing system with authors often submitting to multiple journals, working their way “down the food chain” and studies being reviewed again and again (with the studies often not being modified, and the reviewers’ comments being lost).

7. The time to publication still remains very long, years in some cases.

8. Increasing evidence shows a lack of integrity—and even frank fraud—in some of what is published.

9. Despite the growth of open peer review, much of the process of publishing science is not transparent, meaning that it’s not clear why decisions are made and on what evidence, something that seems antiscientific.

10. Although open access publishing has grown, only a minority of research studies are available to all for free and for reuse.

11. The funders of research often don’t follow up on the research they have funded, failing to ensure that it is published and missing opportunities to evaluate their decisions and whether they are getting value for money.

12. Recent years have seen the appearance of “predatory journals,” “journals” that are nothing but financial scams. Well established scientists are unlikely to fall victim to these journals, but young scientists do—and sometimes they may willingly submit to these journals, knowing that those who read their CVs will not know which are the predatory journals.

13. Some may also see it as a problem that science publishers (commercial and society) continue to make substantial profits from publishing research despite the vast majority of the value of the research ( > 95%) being in the research itself rather than in the processes of the publishers and editors.

14. Publishers continue to make such large profits because universities and funders rely so heavily on publications when assessing researchers. It seems extraordinary that universities and funders should have effectively outsourced such a core function, especially when the “publication game” is so arbitrary.

The historical role of journals in publishing science has been, firstly, to provide some quality assurance, and, secondly, to sort studies, enabling researchers to see the studies important to them. Much evidence shows that journals have failed with quality assurance, and in the age of the internet and ever more sophisticated search engines the second function is redundant—and hardly works anyway when there are tens of thousands of journals.

What matters now are not journals but studies backed up by the data on which they are based.

So if the current way of publishing science is broken what would be a better way? The essence of the idea is that it should be a process run by scientists for scientists making the publication of science much quicker and simpler. Studies with the data behind them would appear not in journals but on databases run by funders, universities, or other research institutions. All studies would be open access, meaning anybody anywhere could see them (and the underlying data), and the whole process would be open from beginning to end—and there would really be no end as debate on the studies and data could continue indefinitely.

A possible process

1. Scientists might be provided with a set of authoring tools to support the process of writing up protocols and studies. Authors would be encouraged to publish protocols.

2. All studies would be required to include data supporting the findings that would be openly accessible in a form that could be used by reviewers and readers. Software that forms the main basis of an article must be open source.

3. Once submitted studies would be screened according to a set of explicit criteria and ruled out if not suitable—perhaps because they were not research studies or didn’t include supporting data. There would be an appeal process.

4. All articles that passed the initial screen would be published (made public) immediately and given a unique identifier, making them citable.

5. Before publication authors would select reviewers according to explicit criteria.

6. Reviewers would be notified as the study was published and asked to comment.

7. Reviewers would be given guidance on what was wanted, and their main function would be to help authors improve their study.

8. Reviewers’ comments would be posted as they arrived with their name and affiliation and would have their own unique identifier, making them citable.

9. Authors and other registered users (with their name and affiliation) would be able to comment on both the study and the reviewer’s reports.

10. Authors would be solely responsible for deciding whether and when to revise their studies. All new versions would be published immediately, be independently citable, and then be sent to the reviewers for further comments.

11. A dynamic and easily understood citation providing the study version number and current reviewer status would be used and updated automatically.

12. All studies, regardless of refereeing status, would remain published on the platform. For those studies that reach an agreed and explicit level of positive refereeing, all study versions, reviewers’ reports, authors’ responses, and any other material would also be made available in PubMed Central and other indexing databases.

F1000 has a process very like this, but this is only one possible process. Other processes might be devised, but the process should be fully transparent, open to all, and run by scientists for scientists.

Such a process would not solve all the current problems with the publishing of science, but it would help with many of them.

  • Full data behind studies would be available to all.
  • It would be easier to publish, meaning that less research would be unpublished.
  • Open real time processes of peer review and commenting by anybody would help improve the quality of what is published, with all studies available, but only some making it over the barrier into PubMed Central.
  • Full availability of data and more space for describing methods should help with reproducibility.
  • Full transparency and posting of protocols might help reduce waste.
  • Authors and reviewers would not waste their time as studies worked their way down the food chain of journals.
  • Publication would be instant.
  • Full transparency, particularly the posting of full data and software used to manipulate the data should help raise the integrity of what is published.
  • The basis for all decisions would be transparent.
  • All studies and the underlying data would be available to all.
  • Funders would be obvious candidates for running the process, meaning they would follow up studies they had funded and ensure their availability to all.
  • There would be no space for predatory journals.
  • Current publishers might offer publishing services but would be unlikely to make the large, undeserved profits associated with journals.
  • The “end of journals publishing science” would at least avoid the unscientific process of evaluators using the impact factor of a journal as a surrogate for the value of a study. It would encourage a concentration on studies rather than journals and the use of other non-bibliometric measures.

The next step should be a series of experiments to explore the use of new processes. Some are already close.

Journals have served science well for 400 years, but there are now clear opportunities to move the publishing of science from the 17th to the 21st century.

Competing interest: RS has long been an advocate of open access and served for seven years on the board of the Public Library of Science. He is now working occasionally as a paid consultant to F1000.

Richard Smith was the editor of The BMJ until 2004. He is now chair of the board of trustees of icddr,b [formerly International Centre for Diarrhoeal Disease Research, Bangladesh], and chair of the board of Patients Know Best. He is also a trustee of C3 Collaborating for Health.

  • costasB

    Excellent piece!
    I totally agree with all your 14 points. As far as your proposal for an open system for evaluate scientific work run by scientists for scientists let me introduce you to an excellent attempt started about a year ago by some colleagues in France. It is called the Self Journal of Science and it is worth a visit: