After 18 months of complex testing and re-testing, Pamela Ronald became certain that she needed to retract two high-profile papers on disease resistance in rice1, 2. The hardest part, says Ronald, a crop scientist at the University of California, Davis, was staying calm — she worried about the implications for current and past lab members and about others spending time replicating potentially faulty work.
The papers had claimed to identify a bacterial protein that could activate an immune response in rice plants with a specific receptor. But when new members of her team were unable to reproduce the results, alarm bells started ringing. Shaken, they decided that the first step was to genotype all the laboratory strains in their collection. Eventually they caught a labelling error: two of the 12 strains thought to lack the protein in question actually lacked a different protein. And the careful backtracking unearthed yet another error: the test, which they had used to verify that this protein could trigger resistance, turned out to be faulty. Despite her distress, throughout the ordeal Ronald was straightforward with journal editors and her colleagues about the likelihood of retractions. She knew that her scientific reputation depended on complete transparency about possible errors. “You just have to set aside emotions and let the scientific process pull you through,” she says.
Oivind Hovland/Ikon Images/Getty
Worldwide, retractions are on the rise: last year alone, scientific journals retracted roughly 500 papers (of more than a million published), compared with fewer than 50 per year in the early 2000s (see Nature 478, 26–28; 2011).
One study3 — in the life sciences — suggests that misconduct, such as plagiarism or falsified data, has been to blame for two-thirds of retractions (see Nature 490, 21; 2012). And behavioural ecologist Daniele Fanelli of the University of Montreal in Canada, who studies the issue, says that at least one-quarter are the result of unfortunate mistakes. The rise in retractions could be because scientists are making more errors, but it could also indicate a growing culture of coming clean on errors. And that, Fanelli says, is a positive trend. “We really need to think more about how to reward retractions that are correcting mistakes — find a way to make them a badge of honour instead of a badge of shame,” he says.
Scientists often treat retractions as dirty secrets. The muted discourse means that the process is often much more confusing, frustrating and embarrassing for researchers, journal editors and universities than it needs to be. Many struggle with the best way to correct the record and with how to salvage viable data. Yet if a retraction is the result of an accident or honest error, it should not be a blot on an otherwise respectable publication record. Scientists and journal editors who have retracted papers say that the process can be handled productively, whether the errors are from contamination, a cell-line mix-up or statistical analyses gone awry. Above all, they say, transparency is key.
Sensing a problem
A decade ago, retractions were far from transparent. “It was not unusual to see 'Paper is retracted' [in the journal] and nothing else,” says Ferric Fang, a microbiologist at the University of Washington in Seattle who also studies retractions. In 2009, realizing that journals lacked policies, the UK-based non-profit organization COPE (Committee on Publication Ethics) published guidelines on how best to correct the scientific record. COPE expects its more than 9,000 journal members to follow the guidelines, which recommend that retraction statements link to the retracted article, be freely available and state who is retracting the article, among other criteria. But perhaps most importantly, says former COPE chair Liz Wager, a retraction notice should include the reason that the retraction was made to clearly distinguish misconduct from honest error. “The retraction has to be complete and honest, and clearly articulate what things are wrong,” says Inder Verma, a molecular biologist at the Salk Institute for Biological Studies in La Jolla, California, and editor-in-chief of the Proceedings of the National Academy of Sciences.
There are several ways to amend a published paper (see 'Retraction guide'). A retraction is reserved for the most severe problems, ones that unravel a paper's conclusions. It is hard to retract a paper and later republish the valid parts, says Arturo Casadevall, editor-in-chief of mBio and a microbiologist at Albert Einstein College of Medicine in Bronx, New York. “Many journals won't allow that.”
Source: COPE; Council of Science Editors
For less serious issues, there are two other options. For the least egregious errors — a mislabelled figure, for example — a correction often suffices. Alternatively, the paper can be partially retracted if the erroneous findings do not invalidate the article's stated conclusions. (COPE discourages partial retractions, however, saying that they “make it difficult for readers to determine the status of the article and which parts may be relied upon”.)
Once alerted to a potential problem, the first step is to identify whether an error has actually occurred and, if so, how. The authors and journal editors then need to decide on the appropriate response. As soon as Ronald realized something was amiss with her original work, she contacted the editors of the journals that had published the papers. “Part of you wants to retract it immediately,” she says. But she knew that she needed to ascertain the facts first.
Her team worked out all the experiments that would be needed to determine the problem. “Everyone had slightly different concerns, but we all wanted to get it right,” she says. Finding the error became all-consuming; other projects languished.
Arturo Casadevall: “The greatest currency we have as scientists is respect.”
And Ronald went a step further. Once she was sure that there was a problem, she contacted colleagues to highlight the issue and gave a public seminar to inform the international community. “I was alarmed that others were trying to build on this work when we couldn't, and I didn't want to waste anyone's time,” she says. Her efforts won her praise both from her colleagues and from the Retraction Watch blog, which reports on scientific retractions and misconduct cases. In a post about Ronald's case, Retraction Watch co-founder Ivan Oransky said that the blog likes “being able to point out when researchers stand up and do the right thing, even at personal cost”.
Deciding whether to issue a full-blown retraction is difficult. Reputations are on the line, and authors may not necessarily agree on how best to amend the mistake. It can take a lot of negotiating with journal editors and colleagues.
That was the experience of Daniel St Johnston, a developmental biologist and director of the Gurdon Institute at the University of Cambridge, UK. He and his colleagues had published two papers4, 5 describing a pathway that they thought was essential for the epithelial cells surrounding a developing egg to maintain their polarity, or orientation, under starvation conditions. However, they later discovered that they had damaged some of the cells when they had dissected the tiny ovaries of the starved fruit flies, resulting in 'false clones' that mimicked the appearance of the cells of interest, but did not have the same mutations.
St Johnston first wrote a new paper highlighting the existence of false clones. Once that paper was accepted in Biology Open6, he sent the manuscript to the two journals that had published the original incorrect papers, Journal of Cell Biology and Developmental Cell, in a bid to find the best way to link the new findings to the original papers. The journal editors advised him to retract the original papers because the main conclusions were no longer valid. Then things got complicated.
One author was not happy about the prospect of losing the data that remained sound from the scientific record. But to retract the papers, all authors had to approve. The researchers and journal editors discussed various options, including a partial retraction, but couldn't reach an agreement. To break the deadlock, St Johnston arranged for the valid data to be published alongside the false-clone findings in Biology Open7, and the retractions went ahead. St Johnston says that he had to figure out the best course as he went along: “There's no obvious mechanism in place to handle the useful, valid content of a retracted paper.” He adds that so far, he has not experienced any career backlash.
The data support what many journal editors advise — that the best way to overcome the stigma associated with a retraction is to come clean with a detailed account of mistakes. A study published last November8 found that authors who self-report mistakes and retract papers accordingly will not lose out on citations to their previous, legitimate research. However, authors who fail to self-report the need for a retraction lose up to 12.5% of citations per year per paper, five years after the retraction, compared with non-retracted papers with similar citation patterns. “Being transparent preserves your reputation as someone honestly seeking truth,” says study author Benjamin Jones, a higher-education researcher at Northwestern University in Evanston, Illinois.
In fact, a clear retraction may end up being worthy of a citation itself. For example, a 2006 retraction9 detailed an author dispute over whether original findings of ferromagnetic behaviour found in a carbon-60 atom remained valid in the face of new measures. “I've found examples of retracted papers that continue to be cited because the retraction is clear, suggesting that people in the field are more discerning than they've been given credit for,” says Fang.
In St Johnston's experience, the more common case is that scientists bury the error — which, he says, harms the scientific community by squandering everyone's time and resources. “I personally didn't want to do that because my reputation is my trademark,” he says. “If I own up to my mistakes, people know that if I get it wrong, I will tell them.”
St Johnston knew that retracting the paper was the right thing to do, but he says that if he had been at an earlier, more precarious stage of his career, he might have thought twice. And he did have serious concerns about the impact it would have on his junior co-authors. “Most of the scientific community doesn't bother distinguishing between an error and something more sinister,” he says.
Indeed, some data suggest that early-career researchers are right to be concerned. Jones has compared how the citation records of eminent and less-well-established authors are affected after a retraction10. “If you are highly established, we see little effect if you are one of the authors on a retracted paper. If you are not well established, you will see a negative effect,” he says.
As a then-assistant professor, Geoffrey Chang had such concerns when he was alerted to a problem with his work, and threw himself into discovering the error. A crystallographer at the University of California, San Diego, he had had a string of successes in the form of awards, grants and high-profile papers starting in 2001, when he was in his late 20s.
When colleagues called his findings into question, Chang pored over his homemade computer program and eventually discovered that he had transposed two data columns. He retracted five papers, and republished the corrected molecular structures in two papers roughly a year after pinpointing the problem. The thorough, transparent re-analysis of the data, he says, corrected the scientific record, which helped to maintain his academic standing. “It was seven years ago,” he says, “and it's still very much a part of me. But it made us a stronger, more careful lab.” Chang has since secured major grants, including from the US National Institutes of Health.
Cases such as Chang's illustrate that retracted papers need not doom a career, even a fledgling one. But navigating the process requires a willingness to deal with policies and distinctions that are not always clear-cut. “The greatest currency we have as scientists is respect — that colleagues respect us and our work,” says Casadevall. “A retraction is survivable — if it's tackled honestly and transparently.”
- Science 326, 850–853 (2009). et al.
- PLoS ONE 6, e29192 (2011). et al.
- Proc. Natl Acad. Sci. USA 109, 17028–17033 (2012). , &
- J. Cell Biol. 177, 387–392 (2007). , , , &
- Dev. Cell 16, 83–92 (2009). , , , &
- Biol. Open 2, 1313–1320 (2013). , &
- Biol. Open 2, 1321–1323 (2013). , , &
- Sci. Rep. 3, 3146 (2013). , , Uzzi B. &
- Nature 440, 707 (2006). et al.
- 2013). , , & The Reverse Matthew Effect: Catastrophe and Consequence in Scientific Teams (National Bureau of Economic Research,