Starting in March, Nature and the monthly Nature Research Journals will experiment with an alternative to their time-tested method of peer review. Instead of the traditional single-blind method, in which reviewers are anonymous but know the authors' identity, authors will be able to choose double-blind peer review, in which both authors and reviewers are unknown to each other.

Alternatives to the traditional single-blind peer-review process are often proposed. Chief among them are double-blind and open peer review, two apparent opposites as in the latter both the authors and reviewers are known to each other. But the reasons cited in favor of these two alternatives are different. On one hand, proponents of open peer review see its transparency as a way to encourage more civil and thoughtful reviewer comments—although others are concerned that it promotes a less critical attitude. On the other hand, advocates of double-blind peer review suggest that it eliminates ad hominem biases, such as those based on gender, seniority, reputation and affiliation (A. Mulligan et al., J. Am. Soc. Inf. Sci. Technol. 64, 132–161, 2013 doi:10.1002/asi.22798). How effectively either method can meet these aspirations while maintaining the necessary level of criticism remains a matter of debate.

Nature experimented with open peer review in 2006, but at the time, despite expressed interest, the uptake from both authors and reviewers was low and the open reviews were not technically substantive. Views about open peer review are probably still evolving, as several journals continue to experiment with variations on this practice. Opinions about double-blind review, however, are remarkably consistent.

In one of the largest studies on peer review—a 2009 international and cross-disciplinary survey of more than 4,000 researchers (J. Am. Soc. Inf. Sci. Technol. 64, 132–161, 2013 doi:10.1002/asi.22798)—76% of respondents indicated that double blind was an effective peer-review system. (By comparison, open and single-blind peer review were considered effective by 20% and 45% of respondents, respectively.) More recently, our own reader survey confirmed the desire to have double-blind peer review as an option. Importantly, this sentiment is widely echoed in conversations with young scientists worldwide. These conversations illustrate a widespread perception that biases based on authorship affect the traditional single-blind peer review, and they have contributed greatly to making us reconsider the proposition.

Nature journals editors have traditionally not embraced double-blind peer review, for several reasons. Some were skeptical of its efficacy, some were concerned about the potential difficulty of recruiting referees, and some still saw it as their responsibility to mitigate the biases that this method tackles. All editors take, and will continue to take, this responsibility seriously by maintaining awareness of any potential predispositions when selecting reviewers and considering their comments. They will also continue to honor reasonable requests to exclude particular reviewers, regardless of the chosen method of peer review. But by definition, unconscious biases may be difficult to identify and to control. Several studies have detected involuntary biases, notably on the basis of gender, in other areas of the scientific enterprise, such as the hiring of laboratory staff, citation habits and speaker lineups at conferences. It is therefore difficult to guarantee a bias-free process.

Since June 2013 (Nat. Nanotechnol. 9, 871–872, 2014), Nature Geoscience and Nature Climate Change have allowed authors to choose between double-blind and single-blind peer review at submission. The uptake of the double-blind method has been much lower than the enthusiasm expressed in surveys would have predicted—no more than a fifth of monthly submissions are going the double-blind route—but no substantial effects on the quality of reviews have been detected. The reactions to the trial among surveyed authors have been sufficiently positive that Nature and the Nature monthly journals have decided to join the experiment. (Nature Communications will join at a later date.)

The responsibility for rendering the manuscript anonymous falls to the authors. Clearly, in some situations, keeping their identity secret will be impossible because of awareness of their work in the specialist community. We also continue to promote policies that support researchers who wish to release data early and to discuss their work with their peers prior to publication, via conferences or preprint servers. Therefore, the double-blind process is optional on all titles. Some will chose it to assuage concerns about biases, others purely by principle.

We will keep this initiative under review, and we welcome comments from authors and reviewers.