For some, the Letters forum is the "progress of science" distilled. The exchange between authors and peers provides a natural check on potential errors in methodology and, in some cases, misconduct. In 2006, Science published 289 letters, according to the ISI database, and Nature published 270 letters. Cell publishes Matters Arising, which are articles submitted by researchers that directly challenge models or findings previously published in the journal.

In the case of Andrew Wakefield's study published in Lancet on the association of autism and the MMR vaccine (Lancet, 351:637-41, 1998), the controversial interpretations put forth by the study's authors triggered a cascade of comments and letters, many of which were responded to by the authors. The open debate eventually resulted in the partial retraction of the paper's suggestion. See Partial Retraction.

The editors of NEJM issued an editor's note after an unusually large response (more than 2000) to a Medical Mystery story laying out the symptoms of an unidentified ailment (article: NEJM , 353:2384, 2005); 69% of the responders correctly identified the unknown condition as scurvy (editor's note: NEJM , 354:419-20, 2006).

In August 2006, editors of Analytical Chemistry issued a warning about the validity of the findings of a study on platinum levels in breast implants, which appeared earlier that year (Anal Chem, 78:2925-33, 2006). Editor Royce Murray responded to several letters published in the journal that stated the study's conclusions could not possibly have been supported by its data. He did some investigation of his own, consulting experts, and then issued the editorial warning (Anal Chem, 78:5233, 2006), having found that the paper's conclusions were subject to some doubt. The journal did not, however, retract the paper, as there wasn't an indication of misconduct.

Corrections submitted by authors can range from a misprint, misspelling, or mislabeling to an incorrect positive or negative sign, or to a completely erroneous figure or statement presented within the paper. The journal Cell issues approximately 100 corrections a year, whereas in 2006 The Proceedings of the National Academy of Sciences ( PNAS ) published 128 corrections. According to Diane Sullenberger, executive editor of PNAS , the number of corrections fluctuates each year but doesn't seem to be on the rise. However, a search of the PNAS database found 69 published retractions in 2000 but only 45 in 1997. The sheer number of publications may open the door for more incidence of error, according to Randy Schekman, editor-in-chief of PNAS , who cited a 4.5% rise in submissions for 2006.

EoC has been an official indexing term in the NLM since 2004 and has been used by several of the major peer-reviewed journals. Editors may issue an EoC in cases where they suspect scientific misconduct. Many editors use the EoC as a sign that the paper, although not retracted, is cast in serious doubt.

In 2005, the editors of NEJM issued an EoC regarding an article on Merck-sponsored Vioxx gastrointestinal outcomes research (VIGOR ) (EoC: NEJM , 353:2813-4, 2005; article: NEJM , 343:1520-8, 2000). The editors had been made aware of three cases of Vioxx-triggered myocardial infarction that were not included in the study, but that had been known to the authors prior to its publication. In their statement, the editors noted that the three excluded studies would have rendered some of the interpretations of the data incorrect.

The authors of the study responded to these accusations and stood by their results, claiming that the cutoff date of their study excluded the myocardial infarctions from being included. NEJM editors didn't agree, but rather than retract the paper completely in response to the authors' justifications, the editors issued an Expression of Concern Reaffirmed. "Articles are retracted when data are discovered to be false," says Jeffrey Drazen, editor of NEJM, in an E-mail. "In this case, the data were accurate, but not complete. We believe that the Expression of Concern Reaffirmed sufficiently communicated what we learned and corrected the scientific record with the publication of the additional data. "As editors, our major concern is the veracity of what we publish. VIGOR was true, but it was not complete. What was needed was to complete the record. The Expression of Concern Reaffirmed completes the record," he says.

In line with maintaining veracity, editors at NEJM have also had occasion to remove an EoC. In 2002, the editors were contacted by a university in Germany that was investigating one of its researchers for scientific misconduct. That researcher had a paper on acute renal failure recently published in NEJM (NEJM, 346:305-10, 2002). As a note to the readers, the journal issued an EoC regarding the original study. At the end of 2003, the German university concluded that the researcher had not committed scientific fraud. In response, NEJM retracted its EoC (NEJM; 349:1965, 2003).

Partial retractions can range in their reach into the community and the severity of the errors retracted. Andrew Wakefield, then a researcher at the National Health Service in the United Kingdom, and colleagues proposed a link between the MMR vaccine and autism in 1998 (Lancet, 351:637-41, 1998). That paper caused a flurry of correspondence as well as years of debate.

In 2004, however, the interpretation that led to the controversy was retracted, although the remainder of the paper was left unretracted (Lancet, 363:750, 2004). Lancet editor Richard Horton wrote an editorial statement to accompany the partial retraction. He referred to evidence that Wakefield and colleagues had withheld information from the editors regarding how children had been selected for the study and who received the results (lawyers used evidence by the study to build a legal case). Horton also said that Wakefield had received nearly $100,000 from a legal group to conduct pilot studies that eventually overlapped with the Lancet paper. Despite these allegations, however, the editors concluded in their statement that while the withheld information was significant, they believed it was not the intention of the authors to deceive. In an E-mail, Horton declined to comment further.

In another case, the authors of a 2002 paper in Molecular and Cellular Biology (Mol Cell Biol, 12:4256-67, 2002.) issued a partial retraction last December, having found that five figures did not accurately represent their supporting data. Therefore, the partial retraction applied to only five figures in the paper, while the "integrity" of the rest of the paper was reiterated (Mol Cell Biol, 21:8215-6, 2006).

PNAS ' Schekman says that such a situation would have been a full retraction in his journal: "We don't have partial retractions." Schekman is currently dealing with a case in which the authors of a paper requested an erratum to change three figures, but PNAS insisted that the paper be fully retracted. (He wouldn't say which paper was affected.) "Even if it's one figure," adds Schekman, "if that fundamentally changes the conclusions of the paper, one has to consider a retraction."

The number of retractions occurring in the major peer-reviewed journals varies among publications and by year. Science and Nature publish the most retractions of the major journals. In the past five years Science issued at least three retractions a year, except for 2002 when 11 retractions were published. Nature has published at least one retraction each year for the past five years, with exception to 2003, in which they published 10 retractions.

Linda Miller, US executive editor for Nature , doesn't see a clear trend in the number of retractions, but she believes that authors are more eager to retract papers now in light of erroneous interpretations or data than their counterparts of 10 to 15 years ago. "They don't want to be associated with allowing something like that to happen," she says.

Other editors disagree. Donald Kennedy, editor-in-chief of Science , does not see a particular eagerness in authors to retract their research. He says many authors will cede errors in their research but stand by the robustness of their main hypotheses. In the notorious Sch?n retraction case, an investigation of scientific misconduct prompted the retraction of sixteen papers in the journals Science and Nature in 2002 and 2003, respectively (Science, 298:30-1, 2002;). The lead researcher, Jan Hendrik Sch?n, stood by his results even after all the retractions had been issued and the investigation by his institution, Bell Labs, had been completed and revealed impossible duplications and erroneous breakthroughs in his research on crystal conductivity.

In December 2006, Geoffrey Chang, at the Scripps Research Institute in La Jolla, Calif., retracted five major papers from Science , PNAS , and The Journal of Molecular Biology (See www.the-scientist.com/news/home/39805/). In Chang's case, the reason for retraction was given as a glitch in the software, which had produced erroneous structures of a bacterial ABC transporter. After nearly a decade of research and five major papers published, new results put out by another group of researchers proved Chang's structures to be impossible. He immediately discovered his software mistake and retracted all five papers.

Although retracting papers can damage a scientist's reputation, Kennedy believes that it shouldn't be a black mark on a researcher's record. "It would really be unfair if [Chang] were discounted for his ability to do good science. No humiliation should be associated with a retraction."

Retracting a paper without the permission of all of the authors is unpleasant for many journal editors, according to Kennedy. In the case that the institution where the research took place has conducted an investigation or called for the retraction of a paper, the journal editors work with the authors as much as possible to get consensus about the retraction. The same goes for a situation in which one or more of the authors of a paper notifies a journal that a paper should be retracted without the knowledge or consent of his or her fellow authors.

Papers retracted without the author's permission can often cause trouble. Two papers identifying the enzyme that produces nitrous oxide in plants, published in 2003 in Cell and 2004 in PNAS , were retracted without the lead author's permission, but by authority of her supervising researcher at the Boyce Thompson Plant Institute at Cornell University (Cell, 113:469-82, 2003; PNAS , 101:8239-44, 2004; retractions in Cell, 119:445, 2004, and PNAS , 101:16081, 2004). Once the researcher, Meena Chandok, had left the lab in 2004, her supervisor told the journals' editors that Chandok had fabricated data and that he and his colleagues could not replicate "critical data." Chandok claimed that the allegations were unfounded; she filed suit against her former supervisor for damaging her reputation and future scientific career.

Chandok, who has been at the University of Maryland in Baltimore since leaving Cornell, would only say in an E-mail that she stands by the veracity and results of the retracted papers. She declined to comment further, citing the ongoing lawsuit.