Who stands to win from double-blind peer review?

Abstract

T he launch of Advances in Regenerative Biology deserves to be celebrated not only for being a promising new journal in an exciting field of the life sciences but also for adopting a notable approach of manuscript evaluation: double-blind peer review. This approach is primarily notable for its exceptional rarity in natural science journals, despite overwhelming support for it from the research community. Here, I summarise what in my view are the key advantages of double-blind peer review over the alternatives and discuss how research and researchers can benefit from its wider implementation. Three main systems of pre-publication peer review have been proposed, which differ with respect to whether the reviewers know the identity of the authors and vice versa: open, single-blind and double-blind. In open peer review both sides know who they are, in the single-blind system only the reviewers know the identities of the authors, whereas in the double-blind approach the identities of both sides are not revealed during the review process. The masking of author and reviewer identities is intended to minimise bias. For example, reviewers may be biased in evaluating the authors’ work based on who they are, where and with whom they work; similarly, authors disgruntled by critical reviews may be biased in their future interactions with the reviewers. There are some who would dispute that ‘serious’ scientists can have any bias at all, be it in their role as reviewers or as authors, but this can hardly be taken as anything more than wishful thinking. Bias is at the core of human nature. Indeed much of basic research methodology is designed to neutralise our inevitable biases in interpreting our data. Moreover, pervasive even if subtle and unconscious bias among researchers with respect to gender, race, country of origin or affiliation has been consistently documented (1 6). As long as research is carried out by humans, bias will be here to stay, and instead of denying it, we should try to reduce it. In doing so, there is no need to reinvent the wheel, as a gold standard approach for minimising bias in situations involving human interactions is well established: double blinding. For example, in a clinical trial setting, neither the doctors nor the patients know who belongs to the placebo control group and who gets the drug, impeding biased data reporting, collection or analysis. One might thus reasonably expect that a similar double-blind approach would be standard fare for the peer-review process. Indeed, in some branches of academia, such as the social sciences, double-blind peer review has become an accepted system perhaps because social scientists are better aware of the ubiquity and pitfalls of human bias. In the natural sciences, however, by far the dominant form of research evaluation is single-blind review, whereas the double-blind practice is exceedingly uncommon. So why is there such a profound rift between the way in which the ‘hard sciences’ are conducted and the way in which they are evaluated? It is certainly not due to insufficient popularity of double-blind review. Researcher surveys regularly reveal double-blind review as the method of research assessment that is most highly favoured by scientists across disciplines (7 11). Yet, journal editors, the group most supportive of the reigning single-blind system, have largely resisted the introduction of double-blind review (12, 13). Different justifications have been put forward for this policy for example, the burden of concealing the authors’ identities or the fact that knowing who the authors are helps to check the novelty of their work and to identify conflicts of interest (12 15). These arguments, however, do not stand up to scrutiny. Electronic submission systems can be designed to automatically exclude the author information in the manuscript received by reviewers; the novelty of the work should be judged against all previous publications, not only those by the authors themselves; conflicts of interest arise from the subject matter of the manuscript, not from its authors and in cases when the conflicts are of a personal nature, double blinding can only reduce their occurrence. The most frequent objection against the double-blind model is that masking the authors’ identities will not always be successful (16, 17). This is undoubtedly true for instance, the reviewer may have seen the data in the manuscript at a meeting. But this cannot argue against double-blind peer review any more than the inability to fully blind some treatment and control groups (e.g. due to characteristic side effects of the drug) can argue against the use of double-blind clinical trials. Importantly, research has shown that in most cases reviewers cannot divine the authors’ identities (16 18). Whenever they manage to do it, double-blind review is reduced to the single-blind form, but this still leaves a majority of cases in which the authors will have a better chance of impartial evaluation. The logic and data behind this thinking should seem clear, but the vast majority of life science and biomedical journals remain unmoved by such arguments and persist in their refusal to implement double-blind review. I therefore wonder if there may be some other, less palatable and thus less widely advertised reasons for this attitude. The ‘never-change-a-winningteam’ type of mindset may be one such reason. But I advances in REGENERATIVE BIOLOGY