From this week, Registered Reports will be among the article formats offered by BMC Biology.
Registered Reports are not a new publication format: they have been an option in psychology and social science journals and for preclinical publications for some years. But it is only more recently that they have begun to be recognized as an attractive option for papers in basic biology.
What are they, and what is attractive about them?
First, what’s the question?
A registered report is submitted as the introduction and rationale for a piece of research, with a detailed description of the proposed methodology and statistical analysis, but no data. Referees are asked, broadly, whether the question addressed is a good one, and the proposed methodology and statistical tests are a valid way to answer it.
If the referees are satisfied on these two counts, the journal makes a commitment to publish the paper once the data are collected irrespective of the answer they deliver. So a negative result is as acceptable as a positive one.
This satisfies an increasing sense that it is as important, in the case of an inherently important question, to know when the answer is no as it is to know when it’s yes. But to date it has been less publishable.
Of course there are provisos.
The original methodology must have been followed, and the results must (according to the referees) provide a valid basis for the conclusions drawn from the study.
There are also provisos to the provisos…There may prove, once the investigation begins, to be good reasons for departing from the proposed methodology: provided that these are submitted to and approved by the journal, they are permitted.
Clearly, too, new questions may suggest themselves.
No more HARKing
To insist that a registered report follow a strict trajectory from the question originally posed to a simple answer would be to betray the very nature of biology, where careful experiment has a tendency to raise new questions and lead to unexpected answers.
With this in mind (and in the light of earlier experience with the format) Registered Reports allow the inclusion of exploratory investigations suggested by the study once it has begun.
But these ad-hoc investigations must be declared as such. And it is also a strict rule that the introduction as originally submitted cannot be changed in the final submission when the data have been collected.
Thus a Registered Report should represent the real path from the formulation of the hypothesis to the eventual conclusions. This is a major departure from common practice for publications in biology, as remarked by Buzz Baum in a recent guest blog.
Most papers no longer tell the story of the research: the introduction is written only after the results are in, and (it seems reasonable to venture) tailored to point in the direction of the conclusions the authors believe their data best uphold. (Hence the neologacronysm – as it were – HARKing – hypothesizing after the results are known)
With the recent concern about the credibility of science, a format that enforces candor must be a good thing.
The mirage of reproducibility
The quest for reproducibility was an important motive for the development of registered reports. If the experimental design is capable of reliable support for the conclusions of a study, and the methodology is described in sufficient detail, then the study should be reproducible.
Well, yes and no. It has been argued eloquently elsewhere – most recently in a comment on the unsatisfactory outcome of the cancer reproducibility project, and perhaps most engagingly by Peter Walter on making a clock – that there are too many unknown variables in most biological experiments to allow them to be repeated precisely.
But enforced transparency certainly helps, and might have saved at least some of the papers whose validity simply couldn’t be tested by the cancer reproducibility project because of insufficient information – and the disappearance of the postdoc who really knew how it was done.
Possibly even more important, subjecting the experimental design to scrutiny before the data are collected allows errors of experimental design and statistical solecisms to be corrected before time, effort and resources have been invested in research that cannot deliver a satisfactory answer.
When we initiated our current policy on standards of reporting, we asked our Editorial Board to comment, and one of their most telling points was that by the time of submission of a paper it was far too late to query the methodology.
It’s not for everyone, of course
Not all biological research fits the Registered Reports format. Probably most won’t. But some, we believe, could benefit greatly from that publication option. And if it encourages greater openness in the presentation of research more generally, that must be good.
We are offering Registered Reports in this belief and in a spirit of experiment: as with any experiment, we don’t know how exactly it will turn out.
Here’s where you submit if you would like to help us find out.