Annals' readers, when encountering an especially controversial or exasperating article, might well ask the question: "Who's responsible?" They may assume that everything published in Annals is official policy of the journal, its editors, or the publisher, the American College of Physicians. If so, they would be wrong.

Because misunderstandings about who is responsible can cause mischief, we provide this brief explanation of responsibility for what is published in Annals. Many of our policies are in common with other peer-reviewed medical journals and some are specific to Annals.

Authors are responsible for the honesty of their work [1]. Editors and their staff cannot verify the accuracy of the data, which would require directly reviewing laboratory procedures and databases, usually at the authors' institution. Authors are also responsible for what is reported in the article. It is their statement, not ours. Everyone else involved in the editorial process—editors, reviewers, statisticians, manuscript editors, and proof readers—is only trying to help the authors report it clearly. Authors should not acquiesce to editorial suggestions they believe are wrong. Because both editors and authors want the best possible article, it is usually possible for them to negotiate a resolution. Finally, authors for Annals, and for many other journals, certify in writing that the work is original and has not been submitted or published elsewhere [2] and that each has contributed sufficiently to be properly considered an author [3]. If authors are uncertain whether they meet these responsibilities, they should share their questions with the editors.

Reviewers advise editors on the strengths and weaknesses of manuscripts being considered for publication and how authors could improve them. They also recommend whether the manuscripts they review should be published. However, they are not directly responsible for deciding whether the manuscript will be published. We hope reviewers regard their part as an important professional responsibility, agree to do it when asked, return reviews promptly, and do a good job. Mostly they do. It is a lot to ask, a time-consuming and anonymous job.

At Annals the editors decide which manuscripts are published, with a great deal of input from reviewers and from 12 deputy and associate editors, who meet weekly to discuss final candidates for publication. Much as we might want to sidestep this responsibility, especially for the 85% of manuscripts we must turn down, the final decision is ours alone. We do not simply count votes on whether a manuscript should be published or not. At the First International Congress on Peer Review in Biomedical Publication in 1990 [4], it was the consensus that reviewers should be consultants, not "referees," in that they advise but should not have final jurisdiction over how the game is played.

Reviewers and editors may come to different conclusions about publishing a given manuscript. Sometimes editors simply disagree with a reviewer's judgment. Beyond that, editors take into account many other factors—competing manuscripts, the length of the queue of articles waiting for publication, the balance of interests editors are trying to represent—that are not known to reviewers. So authors should not be surprised if the reviews seem favorable but their manuscript is turned down anyway.

Reviewers also disagree with each other. In a recent sample of 100 Annals manuscripts, the two reviewers, using a 4-point ordinal scale from "accept" to "reject," gave the same recommendation for only 27 manuscripts; on 8 manuscripts there was complete disagreement (one said "accept" and the other "reject"). Some critics of peer review suggest that lack of agreement means the process is flawed [5]. We, and others [6], disagree. We often choose reviewers to represent different, complementary points of view (for example, a content and a methods expert or a partisan and a challenger of the article's thesis) and there is no reason to expect them to agree. Because we are most interested in reviewers' critical analysis of the manuscript's strengths and weaknesses rather than their recommendations about publication, we may think two reviewers with opposing recommendations both gave excellent advice.

Does the peer review and editing process assure that only perfect manuscripts are published? Certainly not. The process is intended to select which manuscripts are published and then to make those manuscripts better and we believe it works that way. But the end result, the published article, is far from perfect. If we ever doubted this, the doubts have been dispelled when we have seen Annals articles dissected at journal club by residents who, despite being on call the night before, can, with apparent ease, find flaws. Thus, the Annals' "seal of approval" is for the process, not the end result, although we believe the two are related.

If editors have the authority to decide what is and is not published in their journal, to whom are they responsible? As Editors of Annals, we are hired by, and responsible to, the Board of Regents of the American College of Physicians. Many other journals are sponsored by a professional society and have a similar arrangement. The Board has never, in its long history, interfered with day-to-day editorial decisions. Nevertheless, several times a year the Editors report to the Board of Regents, or to its Publication Policy Committee, and discuss editorial policy in general (but not decisions about specific manuscripts). If the sponsoring organization and the editors of its journal regularly disagreed about the general direction of the journal, it is likely the organization would find different editors or the editors would want to work elsewhere. The principle of noninterference with specific editorial decisions, sometimes called "editorial freedom," is held as extremely important by editors [7]. A breach of it can signal the end of a journal [8].

There are other safeguards against editorial irresponsibility. Although often portrayed as secretive despots [9], medical editors actually live in a fish bowl. Their performance is transparent, in every issue, year in and year out. They are also fortunate to have an extraordinary diversity of opinions to inform their decisions. An Editorial Board, made up of thoughtful peers, advises us on the overall direction of the journal's content. All submissions to Annals, with the exception of invited editorials, letters, poems, and quotes, undergo external review, and reviewers are sent an explanation of the final decision. The Letters section offers an opportunity for public criticism [10, 11] and for editors to confess [12].

More fundamentally, the integrity of the peer review and publication process is sustained by a basic level of trust in the professional behavior of all involved, which we believe is well placed. Another safeguard is open disclosure in the journal about how manuscripts are handled [13, 14].

There are no official Annals positions. We decide to publish articles not because we personally agree with them but because we believe they present important information or points of view. We do not speak directly for the American College of Physicians; neither do the authors of individual articles, except when they write official position papers of the College, in which case this is clearly noted. Some other journals, notably Lancet [15] and JAMA, do consider some editorials to be the journal's official position. We do not.

We have no illusions that this editorial will, once and for all, dispel all misunderstandings about responsibility for publications in Annals, but we thought we should put it down as clearly as we could. And by the way, we are responsible for this editorial. We only selected and tried to improve the other articles in this issue.