In the case of proposed publications, an editor sends advance copies of an author's work or ideas to researchers or scholars who are experts in the field (known as "referees" or "reviewers"), normally by e-mail or through a web-based manuscript processing system. Usually, there are two or three referees for a given article.
These referees each return an evaluation of the work to the editor, including noting weaknesses or problems along with suggestions for improvement. Typically, most of the referees' comments are eventually seen by the author; scientific journals observe this convention universally. The editor, usually familiar to the field of the manuscript (although typically not in as much depth as the referees, who are specialists), then evaluates the referees' comments, their own opinion of the manuscript, and the context of the scope of the journal or level of the book and readership, before passing a decision back to the author(s), usually with the referees' comments.
Referees' evaluations usually include an explicit recommendation of what to do with the manuscript or proposal, often chosen from a menu provided by the journal or funding agency. Most recommendations are along the lines of the following:
* to unconditionally accept the manuscript or proposal,
* to accept it in the event that its authors improve it in certain ways,
* to reject it, but encourage revision and invite resubmission,
* to reject it outright.
During this process, the role of the referees is advisory, and the editor is typically under no formal obligation to accept the opinions of the referees. Furthermore, in scientific publication, the referees do not act as a group, do not communicate with each other, and typically are not aware of each other's identities or evaluations. There is usually no requirement that the referees achieve consensus. Thus the group dynamics are substantially different from that of a jury. In situations where the referees disagree substantially about the quality of a work, there are a number of strategies for reaching a decision.
When an editor receives very positive and very negative reviews for the same manuscript, the editor often will solicit one or more additional reviews as a tie-breaker. As another strategy in the case of ties, editors may invite authors to reply to a referee's criticisms and permit a compelling rebuttal to break the tie. If an editor does not feel confident to weigh the persuasiveness of a rebuttal, the editor may solicit a response from the referee who made the original criticism. In rare instances, an editor will convey communications back and forth between authors and a referee, in effect allowing them to debate a point. Even in these cases, however, editors do not allow referees to confer with each other, and the goal of the process is explicitly not to reach consensus or to convince anyone to change their opinions. Some medical journals, however (usually following the open access model), have begun posting on the Internet the pre-publication history of each individual article, from the original submission to reviewers' reports, authors' comments, and revised manuscripts.
Traditionally, reviewers would remain anonymous to the authors, but this standard is slowly changing. In some academic fields, most journals now offer the reviewer the option of remaining anonymous or not, or a referee may opt to sign a review, thereby relinquishing anonymity. Published papers sometimes contain, in the acknowledgments section, thanks to anonymous or named referees who helped improve the paper.